<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Purdy House Publishing on Medium]]></title>
        <description><![CDATA[Stories by Purdy House Publishing on Medium]]></description>
        <link>https://medium.com/@purdyhouse?source=rss-7758e0ccfebf------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*WaUhvg2EElIYj2r0kNdLbg.png</url>
            <title>Stories by Purdy House Publishing on Medium</title>
            <link>https://medium.com/@purdyhouse?source=rss-7758e0ccfebf------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 15:47:12 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@purdyhouse/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The Red Sea Has Been Closed for Two Years. Insurance Explains Why.]]></title>
            <link>https://medium.com/@purdyhouse/the-red-sea-has-been-closed-for-two-years-insurance-explains-why-fed37dc7c420?source=rss-7758e0ccfebf------2</link>
            <guid isPermaLink="false">https://medium.com/p/fed37dc7c420</guid>
            <category><![CDATA[world]]></category>
            <category><![CDATA[history]]></category>
            <category><![CDATA[insurance]]></category>
            <category><![CDATA[economics]]></category>
            <dc:creator><![CDATA[Purdy House Publishing]]></dc:creator>
            <pubDate>Thu, 23 Apr 2026 15:46:01 GMT</pubDate>
            <atom:updated>2026-04-23T15:46:01.505Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9KyD8SFVnwUWMbhgwX94-A.png" /><figcaption>Image Credit: DallE</figcaption></figure><h3>Houthi attacks, war-risk premiums, and why ships still sail around Africa.</h3><p>In November 2023, Houthi fighters in Yemen boarded a cargo ship in the Red Sea, seized its crew, and towed it to port. Over the next 14 months, they attacked more than 100 merchant ships, sank four, and killed at least eight seafarers.</p><p>And they did something that multinational naval coalitions and hundreds of millions of dollars in military spending could not undo.</p><p>They made it too expensive to buy insurance.</p><p>If you work in risk, you already understand what happened next. If you do not, here it is: insurance pricing closed one of the most important waterways on earth. Not a navy. Not a government. The underwriting market looked at the Red Sea, ran the numbers, and the numbers said no.</p><p>The Red Sea is still saying no.</p><h3>The Vulnerability Was Already Proven</h3><p>The Suez Canal carries 12 to 15 percent of global maritime trade. It connects Europe to Asia. In 2023, roughly 26,000 vessels transited the canal, generating a record $10.25 billion in revenue for Egypt.</p><p>If you read this column after the Ever Given piece, you already know how fragile that corridor is. In 2021, a single container ship ran aground and shut down the canal for six days. The insurance claims alone ran into the billions. The world discovered that a meaningful percentage of global trade passed through a channel barely wider than a football field.</p><p>The Houthis discovered the same thing. And they had missiles.</p><p>By early 2024, container ship transits through the Suez Canal had dropped 90 percent. The number of ships fell from 26,434 in 2023 to 13,213 in 2024. Egypt’s canal revenue collapsed 61 percent, from $10.25 billion to $3.99 billion. The Egyptian government reported losing nearly $7 billion in a single year.</p><p>But here is the part that should matter to anyone in the insurance industry: the canal is still open. Egypt is still running convoys. The waterway is not physically blocked.</p><p>The ships are not using it because the insurance says no.</p><h3>How a Premium Becomes a Blockade</h3><p>Before the crisis, war-risk insurance for a Red Sea transit cost under 0.1 percent of a vessel’s hull value. A rounding error.</p><p>By mid-2024, that number had climbed to 0.5 to 1.0 percent of hull value. According to Marsh McLennan, the world’s largest insurance broker, the cost of insuring a ship valued at $100 million jumped from about $300,000 to roughly $1 million per voyage. In the Persian Gulf, the same coverage costs 0.0001 percent. The Red Sea, at peak pricing, was 10,000 times more expensive to insure.</p><p>No government declared a blockade. Insurance pricing created one.</p><p>When the Houthis renewed attacks in July 2025, after months of relative quiet, premiums doubled in a single week. Some underwriters stopped offering coverage at all. Neil Roberts, head of marine and aviation at the Lloyd’s Market Association, said the attacks had “highlighted the need for caution when considering a transit.”</p><p>Three crew members of the Greek cargo ship Eternity C were killed in one of those July attacks. The Houthis held the remaining crew for five months before releasing them in December 2025.</p><p>That is who pays when insurance says no and someone sails anyway.</p><h3>Going Around Africa: The Math</h3><p>When war-risk premiums hit prohibitive levels, every major container line on earth made the same calculation. Maersk. MSC. CMA CGM. Hapag-Lloyd. They rerouted around the Cape of Good Hope.</p><p>The Cape route adds roughly 3,500 nautical miles. That means 10 to 14 extra days of sailing. Fuel consumption increases up to 40 percent, adding $1 million to $2 million per round trip for large container vessels.</p><p>But here is the calculation that drove the decision: a Suez transit costs $300,000 to $700,000 in canal fees. War-risk insurance at crisis pricing adds another $1 million or more. Going around Africa costs more fuel, but you eliminate the toll and the war-risk surcharge.</p><p>When insurance exceeds 0.7 to 1 percent of vessel value, the Cape route is cheaper than the Suez route.</p><p>The United States spent more than $750 million on Operation Rough Rider, bombing over 1,000 Houthi targets across Yemen. The military campaign could not bring the ships back. A ceasefire in Gaza could, briefly, because a ceasefire changed the insurance math.</p><p>That is the hierarchy. Missiles did not close the Red Sea. Premiums did. And premiums do not respond to military strikes. They respond to actuarial data.</p><h3>Ships Are Broadcasting Their Nationality to Survive</h3><p>One detail from this crisis tells you more about how insurance-driven risk works than any whitepaper could.</p><p>Ships transiting the Red Sea have started broadcasting messages on their Automatic Identification System to signal national affiliation. Chinese-affiliated vessels now transmit “ALL CHINESE” and “CHINESE OWNERSHIP” on their AIS. The Houthis are backed by Iran. China is Iran’s largest trading partner.</p><p>The AIS system was designed for navigation and collision avoidance. It has been repurposed as a declaration of political alignment. Ships are not identifying themselves. They are making an argument about why they should not be attacked.</p><p>Underwriters are watching the same signals. If a vessel has even an indirect Israeli connection, coverage becomes harder to find and exponentially more expensive. One analysis estimated that over 15,000 ships, roughly one in six vessels globally, meet the Houthis’ targeting criteria.</p><h3>The Ceasefire the Market Does Not Trust</h3><p>In October 2025, Israel and Hamas reached a ceasefire. The Houthis paused their attacks. Maersk completed successful Suez transits in January 2026. War-risk premiums dropped to 0.2 percent of hull value, the lowest since the crisis began.</p><p>Then the 2026 Iran war started. In February, the Houthis threatened to close Bab al-Mandeb again. In March, they joined the broader conflict alongside Hezbollah and Iraqi militias, firing missiles at Israel.</p><p>As of April 2026, the Houthis have not resumed attacks on commercial shipping. Analysts are not sure why. The International Crisis Group called it a “strategic pause.” Others point to coordination with Tehran, or to the pressure of internal conflicts in southern Yemen.</p><p>The insurance market does not care why. It cares whether. And “whether” is priced by the month, not the headline.</p><p>As Kpler’s analysis noted, after more than 100 attacks, multiple sinkings, and dead seafarers, the market needs months of sustained quiet before adjusting its models. There is a difference between a political ceasefire and an actuarial one. The Red Sea has had the first. It has not yet earned the second.</p><h3>The Pattern You Keep Seeing</h3><p>If you have been reading this series, this pattern should be familiar. Insurance does not regulate by passing laws. It regulates by pricing risk. And when the price gets high enough, the behavior changes, whether the policyholder is a shipping line rerouting around Africa or a school district rewriting its AI governance framework.</p><p>This is the thesis at the center of the work coming out of the Purdy House Institute. Ryan James Purdy, a leader in AI governance, argues in his academic memorandum series that insurance companies have become the de facto regulators of artificial intelligence. Not because they wrote any rules. Because they priced the risk. When an insurer excludes AI-related liability from a commercial policy, it creates a compliance requirement more binding than any government mandate.</p><p>The Red Sea crisis is that dynamic at global scale. A militant group with drones and missiles could not close one of the world’s most critical waterways. But they could make it uninsurable. And uninsurable turned out to mean the same thing as closed.</p><p>If you want to know the full story about how insurance became the most powerful force nobody thinks about, check out <em>The House Always Wins: The Secret History of Insurance</em>, available soon on Amazon.</p><h3>Sources</h3><ul><li>Suez Canal Authority, via Ahram Online (April 2025): Canal revenue fell 61% in 2024, from $10.25 billion to $3.99 billion. Ship traffic dropped 50%. <a href="https://english.ahram.org.eg/NewsContent/3/12/537710/Business/Economy/Suez-Canal-revenues-plummet-to--bln-in--amid-Red-S.aspx">https://english.ahram.org.eg/NewsContent/3/12/537710/Business/Economy/Suez-Canal-revenues-plummet-to--bln-in--amid-Red-S.aspx</a></li><li>Egypt Presidential Statement, via Daily News Egypt (December 2024): Egypt lost nearly $7 billion in Suez Canal revenue in 2024. <a href="https://africa.dailynewsegypt.com/egypt-loses-7-billion-of-suez-canal-revenue-as-red-sea-tensions-disrupt-traffic/">https://africa.dailynewsegypt.com/egypt-loses-7-billion-of-suez-canal-revenue-as-red-sea-tensions-disrupt-traffic/</a></li><li>Marsh McLennan, via Business Insurance (July 2025): War-risk premiums doubled to 1% of vessel value. Insurance on a $100M ship jumped from $300,000 to $1 million per voyage. <a href="https://www.businessinsurance.com/red-sea-shipping-insurance-costs-soar-after-new-attacks/">https://www.businessinsurance.com/red-sea-shipping-insurance-costs-soar-after-new-attacks/</a></li><li>S&amp;P Global (July 2025): War-risk premiums at 0.5–1.0% of hull value in Red Sea versus 0.0001% in Persian Gulf. <a href="https://www.spglobal.com/energy/en/news-research/latest-news/shipping/070825-houthi-attacks-boost-insurance-rates-ship-traffic-not-yet-affected">https://www.spglobal.com/energy/en/news-research/latest-news/shipping/070825-houthi-attacks-boost-insurance-rates-ship-traffic-not-yet-affected</a></li><li>S&amp;P Global (December 2025): Post-ceasefire premiums fell to 0.2% of hull value. <a href="https://www.spglobal.com/energy/en/news-research/latest-news/shipping/120425-maritime-war-risk-premiums-fall-in-red-sea-rise-in-black-sea-amid-changing-security-dynamics">https://www.spglobal.com/energy/en/news-research/latest-news/shipping/120425-maritime-war-risk-premiums-fall-in-red-sea-rise-in-black-sea-amid-changing-security-dynamics</a></li><li>Lloyd’s Market Association, via Arab Weekly (July 2025): Neil Roberts on need for caution in Red Sea transits. <a href="https://thearabweekly.com/red-sea-insurance-costs-soar-after-renewed-houthi-ship-attacks-red-sea">https://thearabweekly.com/red-sea-insurance-costs-soar-after-renewed-houthi-ship-attacks-red-sea</a></li><li>Kpler (November 2025): Insurance market analysis of Red Sea risk and ceasefire implications. <a href="https://www.kpler.com/blog/red-sea-risk-maritime-insurance">https://www.kpler.com/blog/red-sea-risk-maritime-insurance</a></li><li>World Bank, via DocShipper (January 2026): Container ship transits through Suez dropped 90% between December 2023 and March 2024. <a href="https://docshipper.com/shipping/red-sea-crisis-update-route-alternatives-cost-impacts/">https://docshipper.com/shipping/red-sea-crisis-update-route-alternatives-cost-impacts/</a></li><li>FreightAmigo, via Logistics Outlook (December 2025): Cape route adds 10–15 days; fuel consumption up 40%. <a href="https://www.logisticsoutlook.com/ports-shipping/red-sea-crisis-how-fast-will-lines-return-and-at-what-cost">https://www.logisticsoutlook.com/ports-shipping/red-sea-crisis-how-fast-will-lines-return-and-at-what-cost</a></li><li>Ballast Markets (April 2025): Cape route adds $1–2M in fuel per round trip. Cape becomes cheaper when war-risk exceeds 0.7–1%. <a href="https://content.ballastmarkets.com/blog/2025-04-21-suez-vs-cape-2-billion-routing-cost/">https://content.ballastmarkets.com/blog/2025-04-21-suez-vs-cape-2-billion-routing-cost/</a></li><li>ACLED (December 2025): 84% fewer Houthi attacks on commercial shipping in 2025 versus 2024 (7 versus 150). <a href="https://acleddata.com/report/regional-power-struggles-fuel-simmering-tensions-across-red-sea">https://acleddata.com/report/regional-power-struggles-fuel-simmering-tensions-across-red-sea</a></li><li>Congressional Research Service (February 2026): Timeline of Houthi attacks, U.S. military operations, and ceasefires. <a href="https://news.usni.org/2026/02/25/report-to-congress-on-yemen-and-red-sea-security">https://news.usni.org/2026/02/25/report-to-congress-on-yemen-and-red-sea-security</a></li><li>The National (March 2026): Analysis of Houthi strategic pause during Iran war. <a href="https://www.thenationalnews.com/news/mena/2026/03/17/mystery-of-no-houthi-attacks-on-red-sea-ships-three-weeks-into-iran-war/">https://www.thenationalnews.com/news/mena/2026/03/17/mystery-of-no-houthi-attacks-on-red-sea-ships-three-weeks-into-iran-war/</a></li><li>Windward, via S&amp;P Global (July 2025): Over 15,000 ships meet Houthi targeting criteria. <a href="https://www.spglobal.com/energy/en/news-research/latest-news/shipping/070825-houthi-attacks-boost-insurance-rates-ship-traffic-not-yet-affected">https://www.spglobal.com/energy/en/news-research/latest-news/shipping/070825-houthi-attacks-boost-insurance-rates-ship-traffic-not-yet-affected</a></li></ul><h3>About the Author</h3><p>Ryan James Purdy is a leading expert in AI governance and the founder of the Purdy House Institute. He is the author of <em>The House Always Wins: The Secret History of Insurance</em>, available on Amazon. His academic memorandum series argues that insurance companies are the de facto regulators of AI. He has 30 years in education across North America and Europe, and in Online and Blended Learning.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fed37dc7c420" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Your AI Policy Passes Your Board. It Fails Your Insurer.]]></title>
            <link>https://medium.com/@purdyhouse/your-ai-policy-passes-your-board-it-fails-your-insurer-146cc92d3e83?source=rss-7758e0ccfebf------2</link>
            <guid isPermaLink="false">https://medium.com/p/146cc92d3e83</guid>
            <category><![CDATA[insurance]]></category>
            <category><![CDATA[policy]]></category>
            <category><![CDATA[ai-governance]]></category>
            <category><![CDATA[education]]></category>
            <dc:creator><![CDATA[Purdy House Publishing]]></dc:creator>
            <pubDate>Tue, 21 Apr 2026 15:41:01 GMT</pubDate>
            <atom:updated>2026-04-21T15:41:01.348Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gMd82UNc1HhotkpbIxNwXw.png" /><figcaption>Image Credit: DallE</figcaption></figure><h3>Same governance, different evidence demands. The translation problem nobody talks about.</h3><p>Your school has an AI policy. Congratulations. Your board approved it. Your superintendent signed off. It references responsible use and human oversight and all the right principles.</p><p>Now hand it to your insurer.</p><p>Watch what happens.</p><h3>The Problem Nobody Warned You About</h3><p>Here is the situation facing thousands of educational institutions right now. They have done the work. They have written the policy. They have adopted frameworks from UNESCO, the OECD, the NIST AI Risk Management Framework, or their state’s guidance documents. They have committed, on paper, to transparency, fairness, human oversight, and accountability.</p><p>And none of it answers the questions their insurer is about to ask.</p><p>Insurance underwriters do not evaluate whether your institution values fairness. They ask whether you can produce a bias audit log for a specific model. They do not assess whether human oversight is important to your AI philosophy. They ask whether documented protocols exist for human-in-the-loop intervention in specific workflows, and whether logs demonstrate those protocols have been executed. They do not ask whether you have an AI policy. They ask whether you maintain an AI inventory: a complete, timestamped list of every AI system in use, including vendor-provided tools, with risk classifications, data inputs, deployment dates, and named owners.</p><p>The difference is not tone. It is evidentiary logic. Your board asks: “Do we take this seriously?” Your insurer asks: “Prove it. Show me the artifact.”</p><p>This is the translation problem. Your governance exists. The translation does not.</p><h3>One Policy, Four Failures</h3><p>Consider a mid-sized school district that did everything right. They formed an AI task force. They spent months drafting a policy that referenced the NIST AI Risk Management Framework and their state’s guidance. The board approved it unanimously. The superintendent announced it at a staff meeting. Parents received a letter. Everyone felt good.</p><p>Then renewal came. The insurer’s supplemental application asked for an AI asset inventory with risk classifications. It asked for documented incident response procedures for AI-related events. It asked whether the district had conducted bias testing on its student-facing AI tools and could produce the audit records. It asked for named personnel responsible for AI governance decisions, with dates of appointment and evidence of training. The district had none of these artifacts. The policy said all the right things. It produced none of the right evidence. The insurer did not reject the policy. The insurer never read the policy. The insurer read the supplemental application, found it incomplete, and flagged the district for additional underwriting review.</p><p>This is not an edge case. It is the structural condition.</p><p>Different stakeholders evaluate AI governance against entirely different criteria and require evidence in entirely different formats.</p><p>Your board wants to know that you have assigned responsibility and that your policy reflects institutional values. A three-page policy statement answers those questions.</p><p>Your insurer wants to know whether the bias audit log exists, whether the incident response drill was conducted and documented, and whether the vendor risk assessment is on file. Either the artifact exists or it does not.</p><p>Your state regulator may ask you to “ensure meaningful human control” without specifying what documentation would demonstrate compliance. In 28 states, agencies have published or adopted AI guidance for K-12 education, but comprehensiveness and specificity vary dramatically.</p><p>Your vendor’s procurement team wants evidence in yet another format, one that aligns with their checklists, which reference neither the insurer’s questionnaire nor the state’s guidance.</p><p>The result: an institution can hold an AI policy that simultaneously satisfies its board, fails its insurer’s supplemental application, meets state guidance, and stalls vendor procurement.</p><h3>The Deadlines Are Not Theoretical</h3><p>This would be an academic exercise if the deadlines were not real. They are.</p><p>In January 2026, Verisk made new endorsement forms available that allow carriers to exclude generative AI from commercial general liability policies. Three specific ISO forms (CG 40 47, CG 40 48, and CG 35 08) give insurers the option to drop coverage for bodily injury, property damage, or personal injury arising from generative AI. Verisk defines generative AI broadly: any machine-based learning system or model trained on data with the ability to create content. That definition covers chatbots, content generators, AI-assisted coding tools, and dozens of vendor products your staff may already be using. Carrier interest has been strong. Multiple insurers have already filed to adopt the new wordings. Others, including WR Berkley, Cincinnati Financial, and Philadelphia Insurance, have filed their own exclusion language independently. Insurance renewals throughout 2026 are the first cycle where AI governance documentation materially affects coverage terms.</p><p>On June 30, 2026, Colorado’s Consumer Protections for Artificial Intelligence Act takes effect, creating the first binding U.S. state requirement for AI governance in consequential decisions, including education. The law requires impact assessments, risk management programs, and documented governance. Multi-state organizations may treat Colorado’s requirements as a de facto floor.</p><p>On August 2, 2026, the EU AI Act’s high-risk system obligations take effect for educational applications. Institutions with operations in EU member states, or vendors serving EU-based schools, face conformity assessments, technical documentation requirements, human oversight protocols, and transparency disclosures.</p><p>Meanwhile, the federal landscape is not providing clarity. A December 2025 executive order established an AI Litigation Task Force to challenge state laws deemed inconsistent with federal innovation priorities, explicitly naming Colorado’s algorithmic discrimination provisions as a target. The order directs federal agencies to consider conditioning grant funding on states refraining from AI regulation the administration considers onerous. State guidance cannot serve as stable ground for governance investment when the regulatory foundation itself is contested. An institution building its governance program around current state requirements may find those requirements preempted, amended, or abandoned before the documentation has been operationalized.</p><p>This instability does not reduce the need for governance. It increases it. When external standards are fragmented and contested, the burden shifts to institutions to build internally defensible frameworks that can adapt as requirements clarify.</p><h3>We Have Been Here Before</h3><p>If this sounds overwhelming, consider that another industry went through almost exactly the same thing.</p><p>Before SOC 2 existed, cybersecurity compliance was chaos. Every business-to-business relationship involving data required bespoke negotiation over security evidence. Customers sent unique questionnaires, sometimes hundreds of questions in spreadsheet format. Vendors spent thousands of hours answering them with varying levels of rigor. There was no shared language. There was no standard. There was just fragmentation.</p><p>In 2010, the American Institute of Certified Public Accountants created SOC 2, separating security and operational controls from financial controls. A vendor undergoing a SOC 2 audit could satisfy numerous due diligence requirements with a single artifact. Transaction costs dropped. Trust became portable. Before SOC 2, a company like a payroll processor might answer hundreds of unique security questionnaires from different clients every year. After SOC 2, they could point to one independently attested report. The market found a shared language, and the chaos resolved.</p><p>No equivalent to SOC 2 exists for AI governance. ISO 42001 is the closest approximation, but adoption remains limited, costs are prohibitive for most schools, and it does not address education-sector evidence formats. The SOC 2 transition took roughly five years from announcement to widespread market recognition. AI governance standardization may take longer.</p><p>That means we are in the interim period right now. And the interim period is not a brief transition. It is a sustained condition.</p><h3>What This Means</h3><p>Your AI policy is necessary. It is not sufficient. You need governance that produces attestation-grade evidence: named personnel, documented protocols, dated logs, bias audit records, vendor risk assessments, and incident response documentation. Not sentiments. Artifacts.</p><p>You need to understand that the same governance domain produces different documentation demands depending on who is asking. Transparency means one thing to your board, another thing to your insurer, and something else entirely to a Colorado regulator or an EU conformity assessor. Building for one audience and hoping the others accept it is not a strategy. It is a gamble.</p><p>The governance vacuum is not permission to defer. It is a transfer of responsibility from legislatures and regulators to institutional leadership. The institutions that build translation capacity now, while standards have not converged, will be ready when the underwriter calls. The ones that wait will find out at renewal that their policy passes their board and fails their insurer.</p><p>By then, it will be too late to fix.</p><p>This is the same pattern I document throughout <em>The House Always Wins: The Secret History of Insurance</em>, available now on Amazon. Insurance has always forced governance into existence long before governments get around to it. It happened with fire safety. It happened with cybersecurity. It is happening right now with AI.</p><p>The only question is whether your institution will be ready.</p><p><strong>Ryan James Purdy</strong> is a leading expert in AI governance and the author of <em>The House Always Wins: The Secret History of Insurance</em> (Purdy House Publishing), available on Amazon. With 30 years in education across North America and Europe, and in Online and Blended Learning, Ryan advises organizations on AI assurance, liability, and compliance. His research argues that insurance companies, not governments, are the de facto regulators of artificial intelligence. Connect with him on LinkedIn at linkedin.com/in/purdyhouse or visit sellingshovels.org.</p><p><strong>Sources:</strong></p><ul><li>Clark Hill PLC. “Colorado’s AI Law Delayed Until June 2026.” August 28, 2025. <a href="https://www.clarkhill.com/news-events/news/colorados-ai-law-delayed-until-june-2026-what-the-latest-setback-means-for-businesses/">https://www.clarkhill.com/news-events/news/colorados-ai-law-delayed-until-june-2026-what-the-latest-setback-means-for-businesses/</a></li><li>Gridex. “Verisk CG 40 47: What the New AI Exclusions Mean for Your Commercial Clients.” March 2026. <a href="https://gridex.dev/blog/verisk-ai-exclusions/">https://gridex.dev/blog/verisk-ai-exclusions/</a></li><li>IndependentAgent.com. “Verisk to Roll Out New General Liability Exclusions for Generative AI Exposures.” October 2025. <a href="https://www.independentagent.com/vu_resource/verisk-to-roll-out-new-general-liability-exclusions-for-generative-ai-exposures/">https://www.independentagent.com/vu_resource/verisk-to-roll-out-new-general-liability-exclusions-for-generative-ai-exposures/</a></li><li>Purdy, R.J. “The Translation Problem: Evidence Requirements and Stakeholder Variation in Educational AI Governance.” Purdy House Publishing Memorandum №3, December 2025. Available on Zenodo: <a href="https://doi.org/10.5281/zenodo.18085646">https://doi.org/10.5281/zenodo.18085646</a>. Also available on SSRN and Academia.edu.</li><li>STACK Cybersecurity. “Colorado AI Act (SB 24–205) Compliance Guide.” January 2026. <a href="https://stackcyber.com/posts/ai-colorado-laws">https://stackcyber.com/posts/ai-colorado-laws</a></li></ul><p><strong>Tags:</strong> Education, Artificial Intelligence, Insurance, Policy, Governance</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=146cc92d3e83" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Everyone Laughed at the Ship. Nobody Talked About the Sheep.]]></title>
            <link>https://medium.com/@purdyhouse/everyone-laughed-at-the-ship-nobody-talked-about-the-sheep-3c7b2254459d?source=rss-7758e0ccfebf------2</link>
            <guid isPermaLink="false">https://medium.com/p/3c7b2254459d</guid>
            <category><![CDATA[insurance]]></category>
            <category><![CDATA[money]]></category>
            <category><![CDATA[world]]></category>
            <category><![CDATA[history]]></category>
            <category><![CDATA[culture]]></category>
            <dc:creator><![CDATA[Purdy House Publishing]]></dc:creator>
            <pubDate>Sun, 19 Apr 2026 13:46:01 GMT</pubDate>
            <atom:updated>2026-04-19T13:46:01.301Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vx2L7djkTfVwKw6V2rIxHQ.png" /><figcaption>Image Credit: DallE</figcaption></figure><h3>130,000 animals trapped while the internet made memes. The part of the Ever Given story that should scare you.</h3><p>Michael Shah laughed at the memes like everyone else.</p><p>The tiny excavator clawing at the hull of a ship the size of a skyscraper laid on its side. Someone superimposed it onto the moon landing. Someone else turned it into a metaphor for their marriage. For about seventy-two hours in March 2021, the funniest thing in the world was a boat stuck in a ditch. Shah, the CEO of EasyEquipment, a catering supply company in Manchester, probably scrolled past it on his phone between calls. Everyone did.</p><p>He did not yet know his refrigerators were on that ship.</p><h3>The Ditch</h3><p>The Ever Given was, at the time of her grounding, one of the largest container ships on Earth. Four hundred meters long. Longer than the Eiffel Tower laid flat. She drew sixteen meters of water and carried twenty thousand containers stacked in columns that, from a distance, made her look less like a vessel than a floating city block turned on its side.</p><p>On the morning of March 23, 2021, she was the fifth ship in a northbound convoy through the Suez Canal when a sandstorm rolled in, winds gusting above forty knots. Twenty thousand containers stacked on an open deck catch wind like a sail. At 07:40 local time, the Ever Given lost steerage, drifted sideways, and wedged herself between the canal’s banks. Her bow drove into the eastern bank. Her stern lodged into the western bank. The canal, at this section, was not wide enough for another ship to pass.</p><p>Within hours, the queue began to grow. Within days, it would reach over four hundred ships. Twelve percent of global trade passes through the Suez Canal. Nine point six billion dollars in goods stopped moving every single day.</p><p>Shah’s refrigerators were somewhere in those twenty thousand containers. Commercial kitchen equipment, bound for restaurants across the UK that were finally preparing to reopen after a year of pandemic closures. One hundred thousand dollars in cargo. He did not have marine insurance on it.</p><p>And the memes kept coming.</p><h3>The Sheep Nobody Mentioned</h3><p>While Shah waited for updates on his cargo, something far worse was unfolding in the queue behind the Ever Given that almost nobody was talking about.</p><p>At least twenty ships stuck in that growing line were livestock carriers. Most had loaded sheep in Romania, bound for the Middle East. Saudi Arabia is the world’s largest sheep importer by a wide margin, and Ramadan was approaching. Romania had been sending roughly fifty thousand sheep a week.</p><p>The animals were packed into converted car carriers. Ships originally built to transport automobiles, retrofitted with low ceilings and steep ramps for living cargo. There was no veterinary care on board. The waste had nowhere to go. Animals that fell on the steep ramps could be trampled. A 2020 European Commission report had already found that many of these ships were not fit for transporting animals and that oversight was severely lacking. The voyages normally lasted five to eight days. EU regulations required ships to carry twenty-five percent more feed than the scheduled voyage demanded. That margin sounds generous. It is not. Not when the delay stretches to nine days.</p><p>One of the Romanian ships had departed on March 16 with a scheduled arrival in Jordan on March 23, the very day the Ever Given grounded. With the twenty-five percent surplus, the sheep would face food shortages after just two days of delay. That ship would not arrive until April 1 at the earliest.</p><p>By day three of the blockage, the arithmetic was getting difficult across the entire queue. By day five, it was getting impossible.</p><p>Gabi Paun, the EU director of Animals International, tracked eleven Romanian ships stuck in the queue and warned that food supplies had almost certainly run out. The options were brutal: find a port willing to accept undocumented livestock under emergency provisions, turn the ships around for a voyage back to Romania that would take longer than the remaining feed could sustain, or do nothing and let the bodies accumulate. The Suez Canal Authority would not allow dead animals to be disposed of in the canal.</p><p>An estimated 130,000 sheep were trapped in that line.</p><p>On Twitter, someone was photoshopping the excavator onto the Mona Lisa. In the Red Sea, a crew member on a livestock carrier was calculating how many of the animals in his hold would still be alive by the weekend.</p><p>Michael Shah was worried about refrigerators. He had no idea what was happening to the sheep.</p><h3>The Race Against the Moon</h3><p>The Suez Canal Authority called in SMIT Salvage, a Dutch firm that had been pulling ships out of trouble for over a century. Their CEO went on Dutch television and warned it could take weeks.</p><p>The problem was physics. The Ever Given’s bow was buried roughly six meters into the canal’s eastern bank, lodged in rocky soil.</p><p>But there was a deadline, and it was not set by lawyers or governments. It was set by the moon. A supermoon was bringing a spring tide that would raise the water level roughly nineteen inches above the high-water mark on the morning the Ever Given had grounded. If the salvage team could time their effort to the peak tidal flow, the extra water might lift the hull enough for the tugs to pull her free. If they missed the window, the next spring tide was two weeks away. Two more weeks of the canal shut. Twelve percent of global trade, frozen. More sheep dying in the heat.</p><p>On March 29, at two in the morning, the final operation began. The tide was rising. Fourteen tugs strained against the hull. Dredgers dug. Ballast was adjusted.</p><p>At 15:05 local time, after six days, the bow broke free.</p><p>The Ever Given floated.</p><p>The internet celebrated. The memes had a final lap. Everyone moved on.</p><h3>The Letter</h3><p>The Romanian government’s official mortality count for the sheep began only after the ships reached port. Nobody will ever know exactly how many animals died in the Red Sea during those six days. The number does not exist because the counting did not start until the evidence was already in the water.</p><p>Michael Shah’s problems were just beginning.</p><p>His cargo was not released when the ship was freed. In fact, the ship was seized by the Egyptian government on April 13. The Ever Given and its eighteen thousand containers sat anchored in the Great Bitter Lake, a body of water in the Egyptian desert between the northern and southern sections of the canal. In summer, temperatures there regularly exceed forty degrees Celsius. Twenty-five Indian sailors sat on the ship. They were not parties to any dispute. They were not charged with anything. They simply could not go home. Meanwhile, lawyers in Cairo and London negotiated a compensation claim that started at $916 million. Shah’s refrigerators sat in the heat with everything else, and there was nothing he could do about it.</p><p>And then the letter arrived.</p><p>It came from a firm in Liverpool, from a building on the Pier Head waterfront that has housed insurance clerks and adjusters for over a century. It informed Shah that he owed approximately twenty-five thousand dollars in cash before his cargo would be released. Twenty-five thousand dollars he did not have. For a crisis he had not caused. Governed by a legal principle he had never heard of. Contained in a document that carried the name of a coffeehouse that had been serving merchants since 1688.</p><p>Shah did not have marine cargo insurance. Neither did the vast majority of small shippers on the Ever Given. Lloyd’s List estimated that up to ninety percent of affected cargo lacked delay insurance. It was a relatively cheap product. Almost nobody bought it.</p><p>He told CNN that he knew he was going to have to foot the bill to get his stock back. One hundred thousand dollars in refrigerators, held hostage by an ancient law, in a desert lake, on a ship that had been stuck in a ditch while the internet made jokes.</p><p>The law that produced that letter is one of the oldest legal doctrines still in active use anywhere on Earth. It traces back roughly three thousand years. It has never been abolished. It has never been rewritten. And it is the same system that, right now, determines what your business can insure, what your school can do with AI, and what risks your government quietly refuses to cover.</p><p>That story is in Chapter 11 of my book, <em>The House Always Wins: The Secret History of Insurance</em>, available now on Amazon.</p><p>Pick it up. And next time you see a meme about a boat stuck in a ditch, remember the sheep. And remember Michael Shah’s refrigerators.</p><p><strong>Ryan James Purdy</strong> is a leading expert in AI governance and the author of <em>The House Always Wins: The Secret History of Insurance</em> (Purdy House Publishing), available on Amazon. With 30 years in education across North America and Europe, and in Online and Blended Learning, Ryan advises organizations on AI assurance, liability, and compliance. His research argues that insurance companies, not governments, are the de facto regulators of artificial intelligence. Connect with him on LinkedIn at linkedin.com/in/purdyhouse or visit sellingshovels.org.</p><p><strong>Sources:</strong></p><ul><li>Animal Welfare Institute. “Suez Canal Blockage Highlights Plight of Farm Animals at Sea.” AWI Quarterly, Summer 2021. <a href="https://awionline.org/awi-quarterly/summer-2021/suez-canal-blockage-highlights-plight-farm-animals-sea">https://awionline.org/awi-quarterly/summer-2021/suez-canal-blockage-highlights-plight-farm-animals-sea</a></li><li>Animals 24–7. “Sheep &amp; Cattle Stranded at Sea by Suez Canal Blockage Are Moving Again.” March 29, 2021. <a href="https://www.animals24-7.org/2021/03/26/thousands-of-sheep-cattle-stranded-at-sea-by-suez-canal-blockage/">https://www.animals24-7.org/2021/03/26/thousands-of-sheep-cattle-stranded-at-sea-by-suez-canal-blockage/</a></li><li>Bloomberg. “Stuck in Suez: Thousands of Animals Packed Tight on Ships.” March 26, 2021. <a href="https://www.bloomberg.com/news/articles/2021-03-26/stuck-in-suez-thousands-of-animals-packed-tight-in-ship-hulls">https://www.bloomberg.com/news/articles/2021-03-26/stuck-in-suez-thousands-of-animals-packed-tight-in-ship-hulls</a></li><li>CNN Business. “IKEA, Lenovo and many more businesses still have products stuck in the Suez Canal.” June 15, 2021. <a href="https://www.cnn.com/2021/06/15/business/suez-canal-ever-given">https://www.cnn.com/2021/06/15/business/suez-canal-ever-given</a></li><li>The Maritime Executive. “20 Livestock Carriers Caught in Suez Canal.” March 29, 2021. <a href="https://maritime-executive.com/article/20-livestock-carriers-caught-in-suez-canal">https://maritime-executive.com/article/20-livestock-carriers-caught-in-suez-canal</a></li><li>The National. “Suez Canal: Time Running Out for 92,000 Animals Stranded on Ships.” March 2021. <a href="https://www.thenationalnews.com/mena/egypt/suez-canal-time-running-out-for-92-000-animals-stranded-on-ships-1.1192221">https://www.thenationalnews.com/mena/egypt/suez-canal-time-running-out-for-92-000-animals-stranded-on-ships-1.1192221</a></li><li>Wikipedia. “2021 Suez Canal Obstruction.” <a href="https://en.wikipedia.org/wiki/2021_Suez_Canal_obstruction">https://en.wikipedia.org/wiki/2021_Suez_Canal_obstruction</a></li></ul><p><strong>Tags:</strong> History, Money, Culture, Psychology, World</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3c7b2254459d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[An AI Defamed a Solar Company. Google Says It’s Not Responsible.]]></title>
            <link>https://medium.com/@purdyhouse/an-ai-defamed-a-solar-company-google-says-its-not-responsible-4f54ac33c710?source=rss-7758e0ccfebf------2</link>
            <guid isPermaLink="false">https://medium.com/p/4f54ac33c710</guid>
            <category><![CDATA[money]]></category>
            <category><![CDATA[economics]]></category>
            <category><![CDATA[insurance]]></category>
            <category><![CDATA[culture]]></category>
            <dc:creator><![CDATA[Purdy House Publishing]]></dc:creator>
            <pubDate>Sat, 18 Apr 2026 14:01:02 GMT</pubDate>
            <atom:updated>2026-04-18T14:01:02.031Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UegzZLxSFkohpz3eu7V8sg.png" /><figcaption>Image Credit: DallE</figcaption></figure><p>Wolf River Electric is a solar installation company based in Isanti, Minnesota. It was not sued by the Minnesota Attorney General. That fact is not in dispute. Google’s AI said it was anyway.</p><p>Starting sometime in 2024, Google’s AI Overviews feature told anyone who searched for Wolf River Electric that the company was “currently facing a lawsuit from the Minnesota Attorney General due to allegations of deceptive sales practices.” The AI cited four sources. None of those sources said anything of the kind. The Minnesota AG did sue four solar-lending companies in 2024, but Wolf River was not among them. Google’s AI stitched the pieces together and invented a version of events that never happened.</p><p>Wolf River found out the way small businesses usually find out something is wrong: customers started canceling.</p><p>On March 3, 2025, a customer canceled a $39,680 solar contract, citing lawsuits they had found on Google. On March 4, another potential client refused to do business with Wolf River after researching online. On March 5, a customer sent a Wolf River sales representative a screenshot of Google’s AI output as the reason they were walking away — the contract was worth $150,000. On March 11, a nonprofit called to say they were “pulling the plug” on their relationship because of “several lawsuits in the last year with the Attorney General’s Office,” and terminated a solar project and a lighting project totaling over $174,000.</p><p>None of those lawsuits existed.</p><p>By the time Wolf River filed suit against Google in March 2025, it was estimating $24.7 million in losses from 2024 alone. By June 2025, its initial disclosures put total damages at between $110 million and $210 million.</p><p>Google’s response has been clear: it is not responsible for what its AI said.</p><p>When Wolf River filed in Minnesota state court, Google moved the case to federal court, where Section 230 defenses tend to be more favorably received. Section 230 of the Communications Decency Act of 1996 protects internet platforms from liability for content created by third parties. The argument: Google is a platform, the AI summarized content found on the web, so Google is not the publisher of the defamatory statement.</p><p>William McGeveran, dean of the University of Minnesota Law School, articulated both the weakness of that argument and the difficulty of Wolf River’s case in one sentence: “The question in this case is, ‘Who wrote the AI Overview?’ Under current law, I think the answer, pretty clearly, is: ‘Not Google.’”</p><p>In January 2026, Judge Jeffrey Bryan ruled that Google had missed the deadline for filing its removal notice and sent the case back to Minnesota state court. The litigation continues.</p><p>Here is the structural problem the Wolf River case puts on the table.</p><p>Google’s AI cited real sources. Those sources did not say what Google’s AI said they said. Nobody programmed the AI to defame Wolf River Electric. Nobody made a decision, negligent or malicious, to publish a false claim. The system produced text from statistical patterns, the text happened to be false and damaging, and the existing legal framework has no clean answer for who bears the consequence.</p><p>The autocomplete function compounded the harm. When users typed “Wolf River Electric” into Google’s search bar, suggested searches included “Wolf River Electric lawsuit,” “Wolf River Electric lawsuit reddit,” and “Wolf River Electric lawsuit Minnesota settlement.” These suggestions appeared not because lawsuits existed, but because enough people had searched for lawsuits that the algorithm flagged it as a likely next query. The AI’s false output generated enough search activity to feed the autocomplete loop, which generated more searches, which generated more cancellations.</p><p>Reddit users called Wolf River a “possible devil corp.” Competitors cited the Google results in consultations with potential clients. The AG’s office received complaints from Wolf River customers who had followed Google’s AI recommendation to “file a complaint with the Minnesota Attorney General’s office” against a company that had never been accused of anything.</p><p>Wolf River has documented contracts. It has customer names, contract numbers, dates, and dollar amounts. Legal experts have described this as unusually strong evidence of direct causation in a defamation case. If that cannot establish liability, the question of what evidence would be sufficient becomes important. And if it does establish liability, the question of who actually pays becomes equally important.</p><p>Right now, the honest answer to both questions is: unclear.</p><p>Not courts, which are still working out whether Section 230 was designed to protect AI-synthesized output. Not regulators, who have not caught up. Not AI companies, whose terms of service disclaim liability for outputs. And not insurers, most of whom have spent the past two years excluding AI-generated harm from commercial liability policies through underwriting changes most policyholders have not noticed.</p><p>That gap is what <em>The House Always Wins: The Secret History of Insurance</em> is built on. The book makes the case that when courts, regulators, and technology companies cannot answer the liability question, the insurance industry answers it through exclusion clauses and underwriting decisions that determine who operates and who does not. The AI liability question is not waiting for a ruling in Minnesota. It is already being answered in policy language most AI users have never read. The book is available on Amazon.</p><p>Sources:</p><ul><li><strong>Complaint analysis + full exhibits</strong> — Volokh Conspiracy / Reason.com, June 11, 2025 <a href="https://reason.com/volokh/2025/06/11/large-libel-models-small-business-sues-google-claiming-ai-overview-in-searches-hallucinated-attorney-general-lawsuit/">https://reason.com/volokh/2025/06/11/large-libel-models-small-business-sues-google-claiming-ai-overview-in-searches-hallucinated-attorney-general-lawsuit/</a></li><li><strong>Remand ruling (Google missed deadline)</strong> — Volokh Conspiracy / Reason.com, January 12, 2026 <a href="https://reason.com/volokh/2026/01/12/google-missed-key-deadline-in-suit-alleging-googles-ai-libeled-business-court-holds/">https://reason.com/volokh/2026/01/12/google-missed-key-deadline-in-suit-alleging-googles-ai-libeled-business-court-holds/</a></li><li><strong>McGeveran quote + Section 230 analysis</strong> — University of Minnesota Law School, June 13, 2025 <a href="https://law.umn.edu/news/2025-06-13-dean-william-mcgeveran-quoted-star-tribune-about-lawsuit-regarding-ai-false">https://law.umn.edu/news/2025-06-13-dean-william-mcgeveran-quoted-star-tribune-about-lawsuit-regarding-ai-false</a></li><li><strong>Complaint detail + Google response language</strong> — Minnesota Lawyer, June 13, 2025 <a href="https://minnlawyer.com/2025/06/13/wolf-river-electric-google-ai-defamation-lawsuit/">https://minnlawyer.com/2025/06/13/wolf-river-electric-google-ai-defamation-lawsuit/</a></li><li><strong>Case overview + removal to federal court</strong> — Government Technology, June 13, 2025 <a href="https://www.govtech.com/public-safety/minnesota-solar-company-sues-google-over-ai-summary">https://www.govtech.com/public-safety/minnesota-solar-company-sues-google-over-ai-summary</a></li><li><strong>Star Tribune</strong> (original reporting, paywalled — robots.txt blocks fetch, but the article is real and the URL is correct) <a href="https://www.startribune.com/google-ai-overview-lawsuit-defamation-great-river-electric/601371780">https://www.startribune.com/google-ai-overview-lawsuit-defamation-great-river-electric/601371780</a></li><li><em>Ryan James Purdy is a leading expert in AI governance and the founder of Purdy House Publishing and Consulting. He has spent 30 years in education across North America and Europe, and in Online and Blended Learning. His academic memoranda on AI governance and institutional risk are published on Zenodo, SSRN, and Academia.edu.</em></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4f54ac33c710" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Zurich Just Paid $10.9 Billion for the Company That Insures Your Data]]></title>
            <link>https://medium.com/@purdyhouse/zurich-just-paid-10-9-billion-for-the-company-that-insures-your-data-32b2968a3b2b?source=rss-7758e0ccfebf------2</link>
            <guid isPermaLink="false">https://medium.com/p/32b2968a3b2b</guid>
            <category><![CDATA[history]]></category>
            <category><![CDATA[money]]></category>
            <category><![CDATA[economics]]></category>
            <category><![CDATA[insurance]]></category>
            <category><![CDATA[culture]]></category>
            <dc:creator><![CDATA[Purdy House Publishing]]></dc:creator>
            <pubDate>Thu, 16 Apr 2026 13:16:01 GMT</pubDate>
            <atom:updated>2026-04-16T13:16:01.482Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mFA4cIJF4RAeJZ_QiAaQEA.png" /><figcaption>Image Creadit: DallE</figcaption></figure><p>On March 2, 2026, Zurich Insurance Group announced it had agreed to acquire Beazley for $10.9 billion in cash. One of the largest insurance acquisitions in recorded history. The boards of both companies signed off. The deal is expected to close in the second half of 2026.</p><p>Here is the detail that should stop you: Beazley’s cyber gross written premiums had fallen 8% in the first three quarters of 2025. The company was pulling back from the US cyber market, citing falling rates against rising ransomware claims. Its chief underwriting officer said publicly that the market wasn’t reacting to the claims environment the way it should.</p><p>Zurich paid a 59% premium above market value for a company that was retreating from its flagship product line.</p><p>If you work in insurance and that doesn’t demand an explanation, you may want to read it again.</p><p>This was not Zurich’s first attempt. They approached Beazley in June 2025. Beazley said no. Zurich returned on January 4, 2026, at 1,230 pence per share. Rejected January 16. Back again at 1,280 pence. Rejected January 22. Agreement in principle reached in February. Final price on March 2: 1,335 pence per share total, a premium of roughly 59% above Beazley’s volume-weighted average share price in the weeks before the offer period began.</p><p>Five approaches. Three rejections. A price that kept climbing.</p><p>That kind of persistence from one of the world’s largest insurers does not happen over a company they could take or leave.</p><p>What Zurich was buying is not obvious from Beazley’s public profile. Beazley is a specialty insurer based in London, with a Lloyd’s presence and business lines that include marine, aviation, fine art, and professional liability. None of those are what Zurich wanted badly enough to keep coming back.</p><p>The thing Zurich wanted was cyber.</p><p>Beazley wrote its first standalone cyber insurance policy in 2008, sixteen years before this acquisition closed. It has spent that time building underwriting expertise, claims data, actuarial models, and a structured reinsurance program that most competitors cannot replicate from scratch. As of 2024, Beazley held roughly 10% of the global cyber insurance market, placing it at or near the top globally. Its cyber reinsurance structure includes three full catastrophe bonds, representing hundreds of millions in protection against systemic events. This is not a line of business you build in a few years. It is a capability that compounds.</p><p>The global cyber insurance market was worth $22.2 billion in 2025. GlobalData projects it will reach $35.4 billion by 2030. Zurich, with roughly $47 billion in annual gross written premiums across all lines, was not among the top global cyber players. That gap was a strategic problem. Beazley was the fastest way to close it.</p><p>So why was Beazley pulling back from US cyber at the same time Zurich was fighting to buy it?</p><p>Because Beazley’s chief underwriting officer Paul Bantick was right: the market wasn’t pricing the risk correctly. Ransomware claims were up. Premium rates were down. The US book had become unprofitable. Beazley was exercising underwriting discipline, the thing that separates a specialty insurer from a premium collector.</p><p>Zurich CEO Mario Greco described what he was actually acquiring: “exceptional data availability and underwriting expertise.” The policies Beazley wrote over sixteen years generated data. The data is the asset. The retreating US book doesn’t diminish the sixteen years of accumulated intelligence about what digital risk actually costs. If anything, it confirms the discipline that made the data worth having.</p><p>The combined business will write approximately $15 billion in specialty gross written premiums, making it the largest specialty underwriter in the world, headquartered in the UK and built on Beazley’s Lloyd’s foundation.</p><p>For the rest of the market, the question is not whether this deal is interesting. It is what it signals.</p><p>The insurers who built cyber expertise early, who have the claims data and the structured reinsurance programs and the underwriting models that come from sixteen years of pricing digital risk, hold the position Beazley held going into this negotiation. They got to say no three times and still set the final price.</p><p>The insurers who are still treating cyber as an add-on line are not in that position. They are on the other side of the table.</p><p>AI liability is next on that trajectory. The Zurich-Beazley acquisition is a nine-figure bet on that thesis, made by people with nine-figure accountability for being wrong.</p><p>The structural reasons this pattern keeps repeating across every major risk category, fire, marine, aviation, medical malpractice, and now data and AI, are what <em>The House Always Wins: The Secret History of Insurance</em> is built on. It is available on Amazon.</p><ul><li><strong>Zurich official press release, March 2, 2026</strong> — Page loads but renders blank (likely JavaScript-gated). The URL is confirmed real from multiple secondary sources citing it directly. <a href="https://www.zurich.com/media/news-releases/2026/2026-0302-01">https://www.zurich.com/media/news-releases/2026/2026-0302-01</a></li><li><strong>Insurance Journal, March 2, 2026</strong> — Confirmed, loads cleanly. <a href="https://www.insurancejournal.com/news/international/2026/03/02/860121.htm">https://www.insurancejournal.com/news/international/2026/03/02/860121.htm</a></li><li><strong>Reinsurance News, March 4, 2026 (final terms)</strong> — Confirmed, loads cleanly. <a href="https://www.reinsurancene.ws/zurich-and-beazley-reach-agreement-on-8-1bn-transaction/">https://www.reinsurancene.ws/zurich-and-beazley-reach-agreement-on-8-1bn-transaction/</a></li><li><strong>Beazley cyber market forecast (10% market share, reinsurance structure)</strong> — Confirmed, loads cleanly. <a href="https://www.reinsurancene.ws/beazley-forecasts-cyber-insurance-market-to-grow-to-40bn-by-2030/">https://www.reinsurancene.ws/beazley-forecasts-cyber-insurance-market-to-grow-to-40bn-by-2030/</a></li><li><strong>Beazley cyber GWP fell 8%, Bantick quote</strong> — The original paymentweek.com URL is a 404. Replace with PYMNTS, which is confirmed live and cites the same Financial Times reporting: <a href="https://www.pymnts.com/cybersecurity/2025/insurer-beazley-steps-back-from-cyber-market-as-attacks-surge">https://www.pymnts.com/cybersecurity/2025/insurer-beazley-steps-back-from-cyber-market-as-attacks-surge</a></li><li><strong>GlobalData ($22.2B market in 2025, projected $35.4B by 2030)</strong> — Blocked by robots.txt. The figures appear in multiple secondary sources. Best alternative that loads cleanly and cites GlobalData directly: <a href="https://www.lifeinsuranceinternational.com/analyst-comment/zurich-targets-beazley/">https://www.lifeinsuranceinternational.com/analyst-comment/zurich-targets-beazley/</a> (confirmed in original search results)</li></ul><p><em>Ryan James Purdy is a leading expert in AI governance and the founder of the Purdy House Publishing and Consulting. He has spent 30 years in education across North America and Europe, and in Online and Blended Learning. His academic memoranda on AI governance and institutional risk are published on Zenodo, SSRN, and Academia.edu. His latest book: The House Always Wins: The Secret History of Insurance.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=32b2968a3b2b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[$108,000 to Prove You’re Responsible. Most Schools Can’t Afford It.]]></title>
            <link>https://medium.com/@purdyhouse/108-000-to-prove-youre-responsible-most-schools-can-t-afford-it-0d351c610f4a?source=rss-7758e0ccfebf------2</link>
            <guid isPermaLink="false">https://medium.com/p/0d351c610f4a</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[economics]]></category>
            <category><![CDATA[education]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Purdy House Publishing]]></dc:creator>
            <pubDate>Tue, 14 Apr 2026 13:16:01 GMT</pubDate>
            <atom:updated>2026-04-14T13:16:01.336Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4c9WAjTRPmoHZD1Dti1PjA.png" /><figcaption>Image Cerdit: DallE</figcaption></figure><p>The international standard for AI governance in organizations is called ISO/IEC 42001. It is the closest thing the world has to a credible, certifiable framework for proving that your institution manages artificial intelligence responsibly. Regulators in Europe point to it. State legislatures reference it. Insurance underwriters are starting to ask about it.</p><p>Getting certified costs between $30,000 and $108,000, takes six to twelve months, and requires an auditor who understands both AI governance and your specific operational context. Almost no such auditors exist for the education sector.</p><p>The national average teacher salary in the United States is $74,177.</p><p>If your school district wants formal proof that it governs AI responsibly, the bill can run to more than the annual salary of every teacher in the building.</p><h3>The Standard Nobody Can Afford</h3><p>ISO/IEC 42001 was published in December 2023 as the first international standard specifically addressing AI management systems. Its architecture is solid. It addresses policy, roles, vendor governance, human oversight, incident response, and documentation. It aligns with the EU AI Act. It is what a rigorous governance framework looks like.</p><p>It was designed for large enterprises. It was not designed for school districts.</p><p>A 2024 survey of 981 district technology leaders by the Consortium for School Networking found that only 9% had updated governance policies specifically for AI use. Fifty-four percent had no separate AI policy at all.</p><p>That is not indifference. That is a structural problem.</p><p>Schools sit in one of the most precarious positions in the AI governance landscape. They are among the most intensive users of algorithmic systems — admissions screening, learning analytics, automated proctoring, intelligent tutoring, student-facing generative AI tools — and they operate with some of the smallest technology budgets in any professional sector. The populations they serve are among the most legally protected: children whose data is governed by FERPA, COPPA, and an accelerating stack of state-level requirements.</p><p>Certification costs $30,000 to $108,000. The timeline runs six to twelve months. The auditors who understand both the ISO 42001 control structure and the realities of a K-12 environment are, in practice, nearly impossible to find.</p><h3>The Deadline Is Not Moving</h3><p>Two clocks are running.</p><p>The EU AI Act classifies educational AI as high-risk when used for admissions decisions, student assessment, or learning pathway determination. The compliance deadline for those systems is August 2026. Colorado’s AI Consumer Protection Act, which applies to both AI vendors and the schools deploying their tools, takes effect June 30, 2026. Other states are building similar requirements.</p><p>Most schools have not started.</p><p>Think about what that looks like in practice. A district attends an AI literacy workshop and files the certificate. It downloads a policy template from another district’s website, modifies the header, and presents it to the board. It signs a vendor contract for a learning analytics platform without reading the data-sharing provisions. The technology director believes the governance problem has been addressed. The compliance file has documents in it. The board has approved something.</p><p>None of it would survive an audit.</p><h3>Compliance Debt</h3><p>I use the term “compliance debt” to describe what accumulates when institutions cannot access proper governance architecture. It works through two failure modes.</p><p>The first is wasted spend: purchasing compliance measures that satisfy no actual requirement. The workshop certificate, the template policy, the one-page AI principles statement — these satisfy the optics of governance without providing its substance.</p><p>The second failure mode is more dangerous: shortcuts that create undocumented liability. Vendor contracts signed without audit rights. AI tools deployed without documented human oversight. Training conducted without attendance records. The school believes it is covered. The gap between what was believed and what exists becomes visible at the worst possible moment — an audit, a breach, a lawsuit.</p><p>Compliance debt accumulates invisibly. The crisis makes it visible.</p><p>This is not a remote risk. The CoSN data suggests the debt is accumulating across the sector right now. If 54% of districts had no AI policy at all in 2024, and certification takes six to twelve months, and the auditors for the education sector are scarce, then many schools will arrive at the 2026 deadlines with documents that will not hold.</p><h3>What Schools Can Actually Do</h3><p>Formal ISO 42001 certification is out of reach for most educational institutions before regulatory deadlines arrive. The cost is too high. The timeline is too long. The auditors are not there.</p><p>The alternative is ISO-aligned readiness: implementing the deployer-relevant controls from ISO 42001 without pursuing formal certification. Documented AI policy with defined scope. Assigned roles and responsibilities. Vendor governance with contract audit rights. Human oversight mechanisms. Incident response procedures. Documentation that can survive scrutiny.</p><p>These are not enterprise-scale requirements. They are the minimum a school can point to when something goes wrong.</p><p>My research at the Purdy House Institute, published as Memorandum №4 of the PHI Memorandum Series on Zenodo, SSRN, and Academia.edu, argues that ISO-aligned readiness is the practical path for the 2026–2027 regulatory window. Certification can remain a future option. Governance cannot wait.</p><p>The deeper argument — why insurance companies will enforce these standards before governments do, and what that means for every institution trying to navigate the gap — is in <em>The House Always Wins: The Secret History of Insurance</em>, available on Amazon.</p><h3>The Pattern Behind the Problem</h3><p>The ISO 42001 cost barrier reflects a structural market failure that creates what I call the accessibility gap. Enterprise AI governance solutions serve large institutions at price points most schools cannot afford. Low-end awareness resources do not constitute governance. The middle ground — operational governance at accessible price points — is vacant. So schools buy the awareness resources, believe they have addressed the governance problem, and the debt accumulates.</p><p>This is the same pattern the insurance industry has observed in every newly regulated space for three centuries. The standard arrives. The large institutions comply. The small institutions do what they can afford. The gap between the two is where the liability lives.</p><p>The clock is running. Most of the people responsible for school AI governance do not know what the clock is running toward.</p><p><em>Ryan James Purdy is a leading expert in AI governance and AI assurance, with 30 years in education across North America and Europe, and in Online and Blended Learning. He is the founder of the Purdy House Institute and the author of</em> The House Always Wins: The Secret History of Insurance*, available on Amazon.*</p><p><strong>Medium Tags:</strong> Education, Artificial Intelligence, Policy, Economics, Technology</p><p><strong>Sources:</strong></p><ul><li>Purdy, R.J. “ISO/IEC 42001 and the Education Sector: A Critical Analysis.” PHI Memorandum №4, Purdy House Institute, January 2026. <a href="https://doi.org/10.5281/zenodo.18237730">https://doi.org/10.5281/zenodo.18237730</a>. Also available on SSRN and Academia.edu.</li><li>Consortium for School Networking (CoSN). <em>Annual Survey of District Technology Leaders</em>, 2024. <a href="https://www.cosn.org/">https://www.cosn.org</a></li><li>National Education Association. <em>Educator Pay and Student Spending: NEA Rankings and Estimates</em>, 2024–25. <a href="https://www.nea.org/resource-library/educator-pay-and-student-spending-how-does-your-state-rank">https://www.nea.org/resource-library/educator-pay-and-student-spending-how-does-your-state-rank</a></li><li>European Parliament. Regulation (EU) 2024/1689 (EU AI Act). Entered into force August 1, 2024. <a href="https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32024R1689">https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32024R1689</a></li><li>Colorado AI Consumer Protection Act, SB 24–205. Enforcement begins June 30, 2026. <a href="https://leg.colorado.gov/bills/sb24-205">https://leg.colorado.gov/bills/sb24-205</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0d351c610f4a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How a Coffee Shop Became the Most Powerful Institution in Global Finance]]></title>
            <link>https://medium.com/@purdyhouse/how-a-coffee-shop-became-the-most-powerful-institution-in-global-finance-b7260eaed1b5?source=rss-7758e0ccfebf------2</link>
            <guid isPermaLink="false">https://medium.com/p/b7260eaed1b5</guid>
            <category><![CDATA[money]]></category>
            <category><![CDATA[world]]></category>
            <category><![CDATA[culture]]></category>
            <category><![CDATA[history]]></category>
            <dc:creator><![CDATA[Purdy House Publishing]]></dc:creator>
            <pubDate>Sun, 12 Apr 2026 13:46:01 GMT</pubDate>
            <atom:updated>2026-04-12T13:46:01.360Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*f72RE2US8xyCGvehXp59kA.png" /></figure><p>There is a blue plaque on the wall of a Sainsbury’s supermarket in London.</p><p>You can walk past it without noticing. Most people do. It sits at eye level, between the automatic doors and a rack of reduced pastries, and it says that Lloyd’s Coffee House once stood on this site and that from it grew one of the world’s great insurance markets.</p><p>That is the whole sign.</p><p>What the sign does not say is that every catastrophe bond written today, every satellite insured against debris, every pandemic exclusion that will be contested in court traces its institutional lineage back to the lease on a room that is now selling meal deals. The insurance industry controls roughly $7 trillion in annual premiums. It shapes building codes, clinical drug trials, and increasingly, the standards that govern AI deployment. The whole apparatus began here, in a coffee house on Lombard Street, because one man understood something his competitors missed.</p><p>He sold information. The coffee was incidental.</p><h3>The Man Nobody Painted</h3><p>Edward Lloyd was born around 1648, probably in Canterbury, and arrived in London around 1680 with his wife Abigail and not much else. No portrait of him survives. No physical description appears in any record that has come down to us. The man whose name is now on a building on Lime Street, whose institution underwrites satellites and ships and an increasing share of the world’s digital infrastructure, is a historical ghost. Nobody thought to paint him. Nobody wrote down what he looked like.</p><p>He had buried a son before he opened the first coffee house. He would bury others. He lost children the way people lost children in the 1680s: routinely, without recourse, writing the name in the parish register and continuing. Then he went to work in a room where other men calculated the odds on other people’s losses.</p><p>When Lloyd died in 1713, the Flying Post noted only: “On Sunday last Died Mr Lloyd the Coffee-Man in Lombard Street.” Not wealthy. Not titled. A man nobody had thought to paint.</p><p>Sometime between 1685 and 1687, Lloyd had opened on Great Tower Street, in the narrow strip between the Tower of London and the Thames. London had close to two thousand coffee houses by the 1690s. Most of them sold coffee and conversation. Lloyd sold reliable intelligence about ships.</p><p>Which ones had sailed. Which ones had arrived. Which ones were overdue. Which ones had been taken by pirates.</p><h3>The Architecture of Trust</h3><p>Lloyd built a network of correspondents at ports around the world. He posted shipping bulletins on the walls. He installed a raised wooden pulpit from which his staff read aloud the latest intelligence on arrivals, departures, and losses, and the room would go quiet when they did.</p><p>Then he did something that separated him from every other coffee house operator in London.</p><p>He gave ship captains free entry and free drinks.</p><p>A captain just back from Jamaica knew things about weather, routes, port conditions, and piracy that no amount of speculation could replace. Lloyd made sure those captains sat where the underwriters could reach them. The information was the product. The coffee was the reason people stayed long enough to buy it.</p><p>By 1696 he was publishing Lloyd’s News, a printed newspaper three times a week. It died within a year. But Lloyd’s List, founded by a later proprietor in 1734, would run for 279 years. What Lloyd had understood was not a business model. It was a principle: whoever controls the flow of accurate information controls the pricing of risk. Every institution that has ever tried to govern uncertainty has arrived at the same conclusion, from the syndicates at Lombard Street to the technology companies now arguing over who should underwrite AI liability.</p><p>The men who gathered at Lloyd’s coffee house were not operating in a regulated marketplace. There was no regulation. There was no formal marketplace. There was a genius for information and a collection of merchants, investors, retired sea captains, and outright gamblers who came because it was where the action was.</p><p>Some of them were building the modern financial system. Some of them were pirates. That overlap was not an accident. It was structural. And on the same walls where those men posted their shipping notices, other things were posted too.</p><h3>What Was on Those Walls</h3><p>The first documented reference to Lloyd’s coffee house appears in the London Gazette of February 1688: an advertisement offering a reward for five stolen watches, to be reported at Mr. Edward Loyd’s Coffee House in Tower Street.</p><p>A second advertisement appeared in June of the same year.</p><p>I am not going to reproduce it here. What I will tell you is that the same walls that tracked which ships were overdue from Jamaica also carried notices about human beings who had decided they would not be recovered. The same bulletin board. The same institution that became the most powerful risk market in history.</p><p>That history is in the book. So is what it meant for the two centuries of insurance law that followed, and the legal case that eventually forced a court in London to confront what had been insured there.</p><p>I found this while reading through the earliest Lloyd’s records. It stopped me cold.</p><h3>Why It Still Matters</h3><p>I write about AI governance for a living. I am a leading expert in AI assurance, and the argument at the center of my work is not complicated: the institutions that will regulate artificial intelligence are not governments. They are insurers. They always have been.</p><p>Lloyd’s of London did not set out to regulate maritime trade. It set out to price maritime risk. The regulation followed from the pricing, because once you can be excluded from coverage, you have a powerful incentive to change your behavior. AI is no different. The exclusion clauses are already being written. The underwriting standards are already being set. Most of the organizations that will be affected have not read them.</p><p>Edward Lloyd’s coffee house did not become powerful because it sold coffee. It became powerful because it created a room where information could be trusted, and trust could be priced, and the price of trust became the price of everything.</p><p>That is still how it works.</p><p>The full story, including what was posted on those walls and the legal consequences that followed, is in <em>The House Always Wins: The Secret History of Insurance</em>, available now on Amazon.</p><p><em>Ryan James Purdy is a leading expert in AI governance and AI assurance, with 30 years in education across North America and Europe, and in Online and Blended Learning. He is the founder of the Purdy House Institute and the author of</em> The House Always Wins: The Secret History of Insurance*, available on Amazon.*</p><p><strong>Sources:</strong></p><ul><li>Lloyd’s of London institutional history. Lloyd’s of London. <a href="https://www.lloyds.com/about-lloyds/history">https://www.lloyds.com/about-lloyds/history</a></li><li>Lloyd’s List history. Lloyd’s of London. <a href="https://www.lloyds.com/lloyds-list">https://www.lloyds.com/lloyds-list</a></li><li>“Advertisement.” London Gazette, February 1688. №2317. British Newspaper Archive.</li><li>Swiss Re Institute. <em>World Insurance: The Great Pivot East</em> (Sigma №4/2024). <a href="https://www.swissre.com/institute/research/sigma-research/sigma-2024-04.html">https://www.swissre.com/institute/research/sigma-research/sigma-2024-04.html</a></li><li>Raynes, H.E. <em>A History of British Insurance</em>. Pitman, 1948. (Lloyd’s origins, seventeenth-century underwriting practices.)</li></ul><p><strong>Tags:</strong> History, Money, Culture, Economics, World</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b7260eaed1b5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[State Farm Is Being Investigated for Racketeering. Over Roof Claims.]]></title>
            <link>https://medium.com/@purdyhouse/state-farm-is-being-investigated-for-racketeering-over-roof-claims-cd29240f5739?source=rss-7758e0ccfebf------2</link>
            <guid isPermaLink="false">https://medium.com/p/cd29240f5739</guid>
            <category><![CDATA[money]]></category>
            <category><![CDATA[culture]]></category>
            <category><![CDATA[history]]></category>
            <category><![CDATA[economics]]></category>
            <dc:creator><![CDATA[Purdy House Publishing]]></dc:creator>
            <pubDate>Sat, 11 Apr 2026 13:31:01 GMT</pubDate>
            <atom:updated>2026-04-11T13:31:01.761Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Oklahoma’s AG says the insurer predetermined claim outcomes to hit corporate savings targets.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iBOp8Nv7li2AQ2jC8ZxAow.png" /><figcaption>Image Credit: DallE</figcaption></figure><p>In December 2025, Oklahoma Attorney General Gentner Drummond filed a motion to intervene in a private lawsuit against State Farm Fire and Casualty Company. The allegation is not that a claim was mishandled. The allegation is that the company operated a coordinated program to predetermine claim outcomes before adjusters ever set foot on a roof.</p><p>Drummond’s petition accuses State Farm of violating the Oklahoma Consumer Protection Act, the Oklahoma Racketeer-Influenced and Corrupt Organizations Act, and the Oklahoma Deceptive Trade Practices Act. It also alleges civil conspiracy and unjust enrichment. The AG is asking the court for penalties, damages, structural reforms, and the recovery of profits State Farm allegedly obtained through the scheme.</p><p>This is not the first time. In 2018, State Farm agreed to a $250 million settlement in a similar case in Oklahoma. The current litigation is broader in scope, has the weight of the state’s attorney general behind it, and the case is now before the Oklahoma Supreme Court. As of March 2026, more than 900 individual lawsuits have been filed against State Farm in Oklahoma courts, all making similar allegations.</p><h3>What is the Hail Focus Initiative?</h3><p>At the center of the lawsuits is an internal State Farm program that plaintiffs call the “Hail Focus Initiative.” According to court filings, State Farm launched the program around 2020 with the goal of reducing aggregate roof claim payouts across Oklahoma. The lawsuits allege that the company partnered with consulting firm Accenture and engineering company Haag Engineering to develop new claim evaluation methods, first tested in Dallas County, Texas, before expanding to Oklahoma and other states.</p><p>The allegation is specific: State Farm allegedly narrowed the definition of what qualifies as hail damage in ways that do not appear in policyholders’ actual contracts. The AG’s petition states that State Farm “secretly substituted restrictive, extra-contractual standards and used those hidden standards to deny or minimize payment of legitimate covered losses.”</p><p>In plain language: your policy says one thing. The company allegedly evaluated your claim using criteria that exist nowhere in the document you signed.</p><p>State Farm denies any wrongdoing. In legal filings, the company describes the initiative as an effort to improve the accuracy, quality, and consistency of wind and hail claims handling and to address both overpayment and underpayment of claims. The company has stated that it paid more than $1 billion to Oklahoma customers for wind and hail damage over the past two years.</p><h3>The Hursh family</h3><p>The case that triggered the AG’s intervention belongs to Billy and Lacy Hursh of Broken Arrow, Oklahoma. After a hailstorm in October 2023, two contractors told them they needed a full roof replacement. State Farm assessed the roof as being in “fair” condition. Eight months later, a second storm hit. The damage compounded.</p><p>Afraid of what another storm could do, the Hurshes paid out of pocket to replace the roof. Cost: more than $22,000, borrowed against the equity of their home. State Farm is the number one property insurer in Oklahoma, holding roughly 30% of the homeowner market. The Hurshes were not outliers. They were customers of the largest insurer in the state, paying premiums that, according to Drummond, are nearly twice the national average.</p><p>They are not alone. Oklahoma City law firm Whitten Burrage has represented over 125 similar cases, all of which settled. According to attorneys speaking in open court, one case settled for $3 million. The settlements were confidential, which kept the nature of the alleged scheme out of public view until the AG intervened. When State Farm began settling cases one by one after a court granted Whitten Burrage greater access to company records, the pattern became harder to dismiss as isolated disputes.</p><h3>The Senate hearing</h3><p>The practices drew federal scrutiny in May 2025, when Senator Josh Hawley chaired a hearing of the Senate Homeland Security subcommittee on disaster management. Hawley grilled executives from both Allstate and State Farm over what he described as a pattern of adjusters being pressured to alter damage estimates to reduce payouts.</p><p>Two independent adjusters testified under oath that they had been instructed to remove line items from damage estimates and to reclassify hail damage as “wear and tear,” a classification that would trigger an exclusion. One adjuster testified he was taken off a case after the insurer concluded his estimate would be too high.</p><p>A State Farm policyholder from North Carolina, Jacob Vertel, testified that his family, including two young children, was still living in temporary housing months after Hurricane Helene because State Farm repeatedly deemed their heavily damaged home livable. Water was pouring through electrical outlets. The sky was visible through multiple rooms, including the children’s bedrooms. State Farm’s operations vice president apologized to Vertel during the hearing.</p><p>Hawley connected the practices to a confidential McKinsey &amp; Company report from 2007 that proposed a strategy for insurers: delay claims, deny claims, and fight claimants who push back. The McKinsey report was first exposed by Bloomberg Markets in the wake of Hurricane Katrina and has surfaced in insurance litigation ever since. Hawley argued that the pattern described in the McKinsey playbook tracked precisely with the conduct alleged in Oklahoma and beyond.</p><p>He accused the companies of running what he called “a system of institutionalized fraud.” Allstate made $4.6 billion in profits in 2024 while its CEO collected $26 million in compensation. State Farm, as a mutual company, does not report profits in the same way, but its Oklahoma policyholders are paying nearly twice the national average in premiums while the state’s attorney general alleges those premiums are funding a scheme to reduce payouts.</p><p>After the hearing, Hawley sent letters to the CEOs of State Farm, Allstate, Nationwide, Liberty Mutual, Travelers, and American Family Insurance demanding they honor their obligations to policyholders. Days later, severe storms hit eastern Missouri, damaging approximately 5,000 buildings. The timing underscored the stakes.</p><h3>What does the State Farm RICO case mean for insurance governance?</h3><p>For insurance professionals, the State Farm case is a governance story. The company allegedly created an internal standard that diverged from the policy language it sold to customers and used that standard to systematically reduce payouts. That is an operational gap. The same kind of gap this series has been documenting across AI governance, except this time the entity with the gap is the insurer itself.</p><p>In the AI governance space, institutions create policies that promise one thing while their operational practices deliver another. Schools have AI policies but no incident forms, no review calendars, no audit trails. The gap between what a policy says and what an institution can prove it follows is a liability exposure that underwriters, regulators, and now state attorneys general are learning to identify and prosecute.</p><p>In January 2026, Verisk Analytics filed endorsements introducing AI-specific exclusion language into commercial liability policies covering an estimated 82% of the U.S. market. The industry is getting better at identifying governance gaps in its clients. The State Farm case is a reminder that the same scrutiny can be turned inward.</p><p>The entity holding the risk writes the rules. It has always worked this way. But when the rule-writer’s own governance fails, accountability cuts both ways.</p><p>Wind and hail damage accounts for roughly 40% of all homeowner insurance claims nationwide. If the allegations against State Farm are proven, the case will establish a precedent for how attorneys general can intervene when internal claims standards diverge from policy language. For every insurer operating in states with consumer protection acts and RICO statutes, the question is not whether State Farm’s practices were unique. The question is whether the governance infrastructure exists to prove they were not happening elsewhere.</p><p>The full history of how insurers became the invisible regulators of entire industries is in <em>The House Always Wins: The Secret History of Insurance</em> by Ryan James Purdy, a leading expert in AI governance and founder of the Purdy House Institute. Available on Amazon.</p><p><strong>Sources:</strong></p><ol><li>Oklahoma Attorney General Gentner Drummond, motion to intervene in Hursh v. State Farm, December 4, 2025. <a href="https://oklahoma.gov/oag/news/newsroom/2025/december/case-against-state-farm-over-hail-focus-initiative-scheme.html">https://oklahoma.gov/oag/news/newsroom/2025/december/case-against-state-farm-over-hail-focus-initiative-scheme.html</a></li><li>NBC News, “Oklahoma lawsuit alleges secret scheme by State Farm to cheat homeowners,” March 2026. <a href="https://www.nbcnews.com/business/consumer/lawsuit-alleges-state-farm-cheats-homeowners-rcna262814">https://www.nbcnews.com/business/consumer/lawsuit-alleges-state-farm-cheats-homeowners-rcna262814</a></li><li>Oklahoma Watch, “State Farm Dodged Billions in 2018 Settlement, Now Faces Fresh RICO Charges in Oklahoma,” December 18, 2025. <a href="https://oklahomawatch.org/2025/12/18/state-farm-dodged-billions-in-2018-settlement-now-faces-fresh-rico-charges-in-oklahoma/">https://oklahomawatch.org/2025/12/18/state-farm-dodged-billions-in-2018-settlement-now-faces-fresh-rico-charges-in-oklahoma/</a></li><li>Live Insurance News, “Oklahoma: State Farm’s Secret Documents Fight Reaches the State’s Highest Court,” April 2026. <a href="https://www.liveinsurancenews.com/oklahoma-state-farms-documents/8570772/">https://www.liveinsurancenews.com/oklahoma-state-farms-documents/8570772/</a></li><li>Repairer Driven News, “Judge allows Oklahoma AG intervention in State Farm racketeering lawsuit,” January 2, 2026. <a href="https://www.repairerdrivennews.com/2026/01/02/judge-allows-oklahoma-ag-intervention-in-state-farm-racketeering-lawsuit/">https://www.repairerdrivennews.com/2026/01/02/judge-allows-oklahoma-ag-intervention-in-state-farm-racketeering-lawsuit/</a></li><li>S&amp;P Global Market Intelligence, “Allstate, State Farm blasted at Senate hearing over claims practices,” May 14, 2025. <a href="https://www.spglobal.com/market-intelligence/en/news-insights/articles/2025/5/allstate-state-farm-blasted-at-senate-hearing-over-claims-practices-88993421">https://www.spglobal.com/market-intelligence/en/news-insights/articles/2025/5/allstate-state-farm-blasted-at-senate-hearing-over-claims-practices-88993421</a></li><li>Repairer Driven News, “Allstate, State Farm questioned about ‘running a racket’ in Senate subcommittee,” May 16, 2025. <a href="https://www.repairerdrivennews.com/2025/05/16/allstate-state-farm-questioned-about-running-a-racket-and-pattern-of-fraud-in-senate-subcommittee/">https://www.repairerdrivennews.com/2025/05/16/allstate-state-farm-questioned-about-running-a-racket-and-pattern-of-fraud-in-senate-subcommittee/</a></li><li>State Farm Newsroom, “Understanding the Issues in Oklahoma,” March 2026. <a href="https://newsroom.statefarm.com/understanding-the-issues-in-oklahoma/">https://newsroom.statefarm.com/understanding-the-issues-in-oklahoma/</a></li><li>Moneywise, “Oklahoma couple says State Farm offered to cover a fraction of their $22K roof repair,” December 22, 2025. <a href="https://moneywise.com/insurance/home/oklahoma-attorney-general-launches-investigation-into-state-farm-after-more-than-100-complaints-of-denied-delayed-claims-is-this-a-larger-problem">https://moneywise.com/insurance/home/oklahoma-attorney-general-launches-investigation-into-state-farm-after-more-than-100-complaints-of-denied-delayed-claims-is-this-a-larger-problem</a></li></ol><p><strong>Tags:</strong> History, Money, Culture, Society, Economics</p><p><strong>Bio:</strong> Ryan James Purdy is a leading AI governance expert, AI Assurance Advisor, and founder of the Purdy House Institute. He has spent 30 years in education across North America, Europe, and in Online and Blended Learning Environments. His book <em>The House Always Wins: The Secret History of Insurance</em> is available on Amazon. Connect on <a href="https://www.linkedin.com/in/purdyhouse/">LinkedIn</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cd29240f5739" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Everybody Has an AI Policy. Almost Nobody Has a Plan.]]></title>
            <link>https://medium.com/@purdyhouse/everybody-has-an-ai-policy-almost-nobody-has-a-plan-3c1555651f2a?source=rss-7758e0ccfebf------2</link>
            <guid isPermaLink="false">https://medium.com/p/3c1555651f2a</guid>
            <category><![CDATA[history]]></category>
            <category><![CDATA[education]]></category>
            <category><![CDATA[money]]></category>
            <category><![CDATA[society]]></category>
            <category><![CDATA[culture]]></category>
            <dc:creator><![CDATA[Purdy House Publishing]]></dc:creator>
            <pubDate>Thu, 09 Apr 2026 13:41:01 GMT</pubDate>
            <atom:updated>2026-04-09T13:41:01.388Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>The gap between what schools promise about AI and what they can prove.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6M2JlfAaTnwOock_ffJebg.png" /><figcaption>Image Credit: DallE</figcaption></figure><p>Karen Clark spent five years telling insurers their catastrophe models were wrong before Hurricane Andrew proved it. She had the math. She had the physics. What she did not have was an industry willing to look at numbers that contradicted three centuries of institutional habit.</p><p>The same pattern is running through education right now, except nobody has built the model yet.</p><p>Schools across the world have AI policies. UNESCO adopted its AI ethics recommendation in 2021, ratified by 194 member states. The EU AI Act classifies educational AI as high-risk, with full compliance obligations arriving in August 2026. The OECD AI Principles have been endorsed by over fifty countries. In the United States, all 50 states are now considering AI-related legislation, building on 1,208 bills introduced in 2025. Ohio requires every public school to have a formal AI policy by July 2026. California, Florida, and Illinois have their own mandates in various stages.</p><p>The volume of governance is staggering. The infrastructure to implement any of it is not.</p><h3>What is the operational gap in AI governance?</h3><p>Ryan James Purdy, a leading voice in AI governance and founder of the Purdy House Institute, mapped this gap in the Institute’s first memorandum. The analysis examined frameworks from UNESCO, the OECD, UNICEF, the European Union, and national governments from the United States to Singapore, and it established a taxonomy for understanding what is actually happening beneath the surface of all this activity.</p><p>There are four categories of AI governance framework. Aspirational frameworks articulate principles. Regulatory frameworks establish legal requirements. Sector guidance provides education-specific recommendations. Operational frameworks supply implementation infrastructure.</p><p>The first three categories have proliferated. The fourth barely exists.</p><p>That distinction is the whole problem. An aspirational framework tells a school that AI systems must avoid discriminatory bias. A regulatory framework makes that a legal requirement, with penalties reaching 35 million euros under the EU AI Act. Sector guidance tells schools to consider the implications.</p><p>None of them tells a school what to do on Monday morning when a teacher reports that an AI tutoring tool gave a student racially biased feedback. There is no incident form. There is no escalation path. There is no review calendar that would have flagged the tool before it reached the classroom. There is no audit trail proving the school evaluated the tool’s bias risk before deployment. The principal has a policy on the wall and nothing in the filing cabinet.</p><p>This is not an edge case. The UNESCO Recommendation requires avoiding discriminatory bias but provides no bias assessment methodology, no documentation formats, no remediation procedures. The OECD Principles require transparency but provide no disclosure templates, no challenge procedures for parents contesting AI-influenced outcomes. UNICEF recommends restricting children under thirteen from generative AI platforms but does not specify how to configure the identity management systems that would enforce it. Three of the most respected governance bodies on Earth, three clear requirements, zero operational tools to implement any of them.</p><p>Every single framework the memorandum examined showed the same pattern. Principles proliferate. Procedural templates, governance calendars, and evidence artifacts are virtually absent.</p><h3>Why the gap is structural</h3><p>This is not a criticism of the organizations producing frameworks. UNESCO is not supposed to write intake forms. The OECD is not in the business of creating quarterly review calendars for individual schools. The EU AI Act follows the European Union’s “New Legislative Framework” approach: it establishes essential requirements while deliberately delegating technical implementation to standards bodies and market actors. That delegation assumes someone else will build the operational layer. In education, nobody has.</p><p>The gap is structural, not incidental. International bodies build consensus. Regulators create enforceable requirements. Sector associations provide advisory guidance. None of them occupies the role necessary to provide the templates, role specifications, configuration guides, and audit documentation that constitute verifiable governance.</p><p>Consider what the EU AI Act actually requires of a school deploying an AI-powered assessment tool. Article 9 mandates a risk management system that runs continuously throughout the AI system’s lifecycle. Article 14 requires human oversight measures that enable staff to fully understand the system’s capabilities, properly interpret its outputs, and override its decisions. Article 10 requires that training data be relevant, representative, and free of errors, which translates into vendor assessment obligations for schools that purchase rather than build AI tools. These are legal requirements with real penalties attached. And for every one of them, the school must build the procedural architecture from scratch: the risk checklists, the severity matrices, the vendor questionnaires, the override protocols, the documentation formats. The Act tells them what they must achieve. It does not tell them how.</p><p>In North America, the regulatory landscape is even more fragmented. FERPA was written in 1974. COPPA was written in 1998. Neither was designed for AI governance. A school trying to determine whether an AI vendor qualifies for FERPA’s “school official” exception, or whether AI-generated inferences about a student constitute a new educational record, is interpreting half-century-old law in contexts it never anticipated, with minimal operational guidance for the AI-specific questions that matter most.</p><p>Picture a district superintendent in Ohio who has just signed a contract with an AI-powered math tutoring platform. The platform collects student performance data, generates learning profiles, and adapts instruction based on behavioral patterns. The superintendent has an AI policy. It says the district is committed to responsible AI use, equity, transparency, and student privacy. It says all of these things because every AI policy says all of these things.</p><p>Now the questions start. Does the platform’s data collection trigger COPPA obligations for students under thirteen? Does the learning profile it generates constitute a new educational record under FERPA? Who in the district is responsible for reviewing the platform’s bias metrics, and how often? What happens if a parent demands to know why the AI recommended their child for remedial placement? The policy answers none of these questions, because the policy was written from principles, not from procedures. The superintendent is standing in the gap between what governance promises and what governance can prove.</p><h3>Why is AI governance becoming an insurance issue?</h3><p>The gap mattered less when AI governance was aspirational. It matters now because it is being priced.</p><p>In January 2026, Verisk Analytics, the company that acquired Karen Clark’s original catastrophe modeling firm AIR, filed endorsements introducing AI-specific exclusion language into commercial liability policies covering an estimated 82% of the U.S. market. The same institutional architecture that Clark built to help insurers see risk more clearly is now being used to decide which risks to walk away from. Insurance underwriters are beginning to ask institutions whether they can demonstrate AI governance, not just assert it. The distinction between “we have a policy” and “we can prove we follow it” is the same distinction that separated insurers who survived Hurricane Andrew from the ones that did not.</p><p>A secondary school in the European Union deploying AI-powered assessment tools must now demonstrate compliance with the AI Act’s human oversight requirements, the GDPR’s data protection obligations, and applicable national education law. A North American district implementing generative AI must navigate FERPA, COPPA, and state-level mandates that multiply by the month. The compliance surface has expanded dramatically. Institutional capacity to address it has not.</p><p>Before catastrophe models, insurers priced risk from experience and assumed the past was a reliable guide. After catastrophe models, they priced from physics and discovered their reserves were a fraction of what they needed. The education sector is in the Old Library right now. It is sitting in rows of dark suits, looking at policies built from principles, and assuming the past is a reliable guide to AI risk. The math exists. The room is not ready to hear the number.</p><p>The full analysis of the operational gap across all major AI governance frameworks, including the forcing functions that are making this urgent, is available in the Purdy House Institute Memorandum Series on <a href="https://doi.org/10.5281/zenodo.18007387">Zenodo</a>, <a href="https://papers.ssrn.com/sol3/cf_dev/AbsByAuth.cfm?per_id=7625798">SSRN</a>, and <a href="https://independent.academia.edu/RyanJamesPurdy">Academia.edu</a>.</p><p>If you want to know the full story about how things got this messed up, check out <em>The House Always Wins: The Secret History of Insurance</em>, available on Amazon.</p><p><strong>Sources:</strong></p><ol><li>UNESCO, Recommendation on the Ethics of Artificial Intelligence (2021). Ratified by 194 member states. <a href="https://www.unesco.org/en/artificial-intelligence/recommendation-ethics">https://www.unesco.org/en/artificial-intelligence/recommendation-ethics</a></li><li>European Union AI Act (Regulation (EU) 2024/1689), classifying educational AI as high-risk under Annex III. High-risk obligations apply August 2, 2026. <a href="https://artificialintelligenceact.eu/high-level-summary/">https://artificialintelligenceact.eu/high-level-summary/</a></li><li>OECD AI Principles (2019, updated 2024). Endorsed by 50+ countries. <a href="https://oecd.ai/en/ai-principles">https://oecd.ai/en/ai-principles</a></li><li>UNICEF, Policy Guidance on AI for Children (2021). <a href="https://www.unicef.org/globalinsight/reports/policy-guidance-ai-children">https://www.unicef.org/globalinsight/reports/policy-guidance-ai-children</a></li><li>All 50 U.S. states considering AI-related bills; 1,208 bills introduced in 2025. Education Commission of the States / AI Governance and Insurance Headlines, January-March 2026.</li><li>Ohio House Bill 96, requiring public school AI policies by July 1, 2026. Ohio Department of Education.</li><li>Verisk Analytics ISO Circular LI-CF-2024–175, introducing AI-specific liability exclusion endorsements (January 2026).</li><li>EU AI Act Articles 9 (risk management), 10 (data governance), 14 (human oversight). Regulation (EU) 2024/1689. <a href="https://artificialintelligenceact.eu/">https://artificialintelligenceact.eu/</a></li><li>FERPA (20 U.S.C. 1232g, 1974); COPPA (15 U.S.C. 6501–6506, 1998).</li><li>Purdy, R.J. (2025). “The Operational Gap in Educational AI Governance: A Landscape Analysis of International Frameworks.” Purdy House Institute Memorandum №1. Zenodo: <a href="https://doi.org/10.5281/zenodo.18007387">https://doi.org/10.5281/zenodo.18007387</a> | SSRN: <a href="https://papers.ssrn.com/sol3/cf_dev/AbsByAuth.cfm?per_id=7625798">https://papers.ssrn.com/sol3/cf_dev/AbsByAuth.cfm?per_id=7625798</a> | Academia.edu: <a href="https://independent.academia.edu/RyanJamesPurdy">https://independent.academia.edu/RyanJamesPurdy</a></li></ol><p><strong>Bio:</strong> Ryan James Purdy is a leading AI governance expert, AI Assurance Advisor, and founder of the Purdy House Institute. He has spent 30 years in education across North America, Europe, and in Online and Blended Learning Environments. His book <em>The House Always Wins: The Secret History of Insurance</em> is available on Amazon. Connect on <a href="https://www.linkedin.com/in/purdyhouse/">LinkedIn</a>.</p><p><strong>Tags:</strong> History, Money, Culture, Education, Society</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3c1555651f2a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[She Dragged a 20-Pound Computer Up 4 Flights of Stairs and Changed the World]]></title>
            <link>https://medium.com/@purdyhouse/she-dragged-a-20-pound-computer-up-4-flights-of-stairs-and-changed-the-world-13f9a7710548?source=rss-7758e0ccfebf------2</link>
            <guid isPermaLink="false">https://medium.com/p/13f9a7710548</guid>
            <category><![CDATA[education]]></category>
            <category><![CDATA[society]]></category>
            <category><![CDATA[culture]]></category>
            <category><![CDATA[insurance]]></category>
            <category><![CDATA[money]]></category>
            <dc:creator><![CDATA[Purdy House Publishing]]></dc:creator>
            <pubDate>Tue, 07 Apr 2026 14:51:01 GMT</pubDate>
            <atom:updated>2026-04-07T14:51:01.664Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Karen Clark built the first catastrophe model. The industry laughed. Then Hurricane Andrew hit.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JDCF1C_BrtXX-UDaugTk5w.png" /><figcaption>Image Credit DallE</figcaption></figure><p>The Lloyd’s building at 1 Lime Street had been open for less than a year. Richard Rogers had designed it inside out, the pipes, the ducts, the staircases, all bolted to the exterior like an anatomy lesson performed on the City of London. Queen Elizabeth had opened it in November 1986. The architecture critics called it radical. The people who worked inside it were not.</p><p>The Old Library survived the move. It had been part of the original 1928 building, dark wood polished to the colour of cold tea, portraits of previous chairmen in gilt frames, tall windows letting in the thin grey light of a London afternoon. The room smelled like old paper and three centuries of institutional certainty. The carpet was thick enough to muffle footsteps, which meant the room was designed, whether anyone intended it or not, to make you feel like you were intruding.</p><p>One hundred men were waiting.</p><p>They were catastrophe reinsurance underwriters. They priced hurricanes, earthquakes, the large and terrible events that could destroy a portfolio in an afternoon. And they priced them the way their predecessors had, and their predecessors before that: experience, intuition, the memory of the last bad year, and a margin loaded on top. The past was the guide. It had always worked. The room itself was the evidence.</p><p>That is the pattern in insurance. The entity holding the risk resists better information until it cannot afford to anymore. It was true in 1688. It was true in that room in 1987. It is true right now.</p><p>The door opened at the back of the room.</p><p>Karen Clark was American. She was female. She was seven months pregnant. And she was carrying a portable computer that weighed roughly twenty pounds, a machine most of the men in this room had never touched. Nobody stood up. Nobody offered to help.</p><p>She walked the length of the Old Library with the computer pulling at her arms and the weight of the pregnancy shifting her centre of gravity forward, past the portraits and the dark wood and the hundred pairs of eyes that tracked her the way you track something unexpected at a formal dinner, with interest and the faintest edge of alarm.</p><p>She reached the front. She set the computer on the table. Twenty pounds of plastic landed on old English oak with a thud that did not belong in this room. The screen was small and grey-green, the colour of an unripe lime. She pressed the power button. The cursor blinked in the upper left corner while three hundred years of accumulated underwriting wisdom sat in rows of dark suits and watched a pregnant American woman wait for a computer to start.</p><p>Clark had come from Boston to tell them that everything they believed about catastrophic risk was wrong.</p><h3>What is a catastrophe model?</h3><p>Before Clark, the insurance industry priced catastrophic risk by looking backward. You took the last thirty or fifty years of hurricane landfalls. You counted the storms. You averaged the losses. You loaded a margin. If the last decade had been quiet, the rate dropped. It was the actuarial equivalent of driving by looking in the rearview mirror.</p><p>The method had a specific blindness. A Category 5 hurricane had not made landfall in south Florida within living memory. The experience tables said it was unlikely because it had not happened recently. The physics said it was overdue.</p><p>Clark had spent years in the research department at Commercial Union Assurance in Boston, where one of her first assignments was assessing whether the company had too much coastal exposure. What she found was not one insurer’s problem. It was a flaw in the way every insurer on Earth priced catastrophic risk.</p><p>Her solution was to stop looking backward. Use physics instead of history. Simulate thousands of possible hurricanes using what meteorologists and structural engineers actually knew about how storms form, intensify, and destroy buildings. Calculate the probable maximum loss from simulation, not from the last bad year. The shift was the difference between “what happened” and “what could happen.”</p><p>In 1986, she published “A Formal Approach to Catastrophe Risk Assessment and Management” in the <em>Proceedings of the Casualty Actuarial Society</em>. It was the founding document of a field that did not yet have a name. The model flowchart on page 5 has remained virtually unchanged for nearly four decades. In 1987, she founded Applied Insurance Research in Boston. The industry was not interested.</p><h3>The silence</h3><p>Back in the Old Library, Clark showed them the model’s output. The probable maximum loss for a major Florida landfall, calculated from physics, not from memory.</p><p>The number was larger than anything the men in this room had priced for. Larger by multiples. It said the catastrophe they believed was unlikely was not only possible but, when it arrived, would cost several times more than the industry was prepared to pay.</p><p>A man in the second row uncrossed his arms and crossed them again. Someone near the back checked his watch.</p><p>Silence. And little interest.</p><p>She packed up the computer. Twenty pounds in her arms again. She carried it back across the Old Library, past the portraits and the dark wood and the hundred men who were already turning to each other with the comfortable certainty that what they had just seen was clever but unnecessary. The door closed behind her.</p><p>Outside, the Rogers building’s exposed ducts gleamed in the London grey. The most radical building in the City of London, housing the most conservative professionals in the City of London, and a pregnant American woman walking out with a computer full of numbers they would need in five years and did not want today.</p><p>Clark had approximately thirty clients when the storm finally came. Mostly London reinsurance companies, forward thinkers she had found after that silence. The conventional worst-case estimate for a major Florida hurricane sat between four and six billion dollars. Lloyd’s put the figure at six billion, tops.</p><p>On August 24, 1992, Hurricane Andrew made landfall just south of Miami. Category 5. Winds at 165 miles per hour.</p><p>Clark’s model had an answer before anyone else did. What that answer was, what the industry refused to believe even as the fax machines started printing, and the five-quid bet one London reinsurer lost because of it, is one of the best stories in the history of risk. It is Chapter 8 of <em>The House Always Wins: The Secret History of Insurance</em>, coming soon.</p><p>What matters here is the pattern. The industry looked backward, priced from experience, and ignored better information until reality arrived and proved the numbers wrong. That pattern is not historical. It is happening right now with AI.</p><p>In January 2026, Verisk Analytics, the company that acquired Clark’s original firm AIR, filed endorsements introducing AI-specific exclusion language into commercial liability policies covering an estimated 82% of the U.S. market. The tool Clark built to help the industry see risk more clearly became the institutional architecture that decides which risks to walk away from. The entity holding the risk writes the rules. It has always worked this way. (For the full analysis of the Verisk exclusions, see the Purdy House Institute Memorandum Series, available on <a href="https://doi.org/10.5281/zenodo.18007387">Zenodo</a>, <a href="https://papers.ssrn.com/sol3/cf_dev/AbsByAuth.cfm?per_id=7625798">SSRN</a>, and <a href="https://independent.academia.edu/RyanJamesPurdy">Academia.edu</a>.)</p><p>If you want to know the full story about how things got this messed up, check out <em>The House Always Wins: The Secret History of Insurance</em>, coming soon.</p><p><strong>Sources:</strong></p><ol><li>Karen Clark biography and career history. Illinois State University, Katie School Innovation and Leadership Award, 2018. <a href="https://business.illinoisstate.edu/centers/katie/hall-of-fame/innovation-awards/clark/">https://business.illinoisstate.edu/centers/katie/hall-of-fame/innovation-awards/clark/</a></li><li>Karen Clark interview on Hurricane Andrew’s 30th anniversary. Insurance Journal, June 2022. <a href="https://www.insurancejournal.com/news/national/2022/06/01/669875.htm">https://www.insurancejournal.com/news/national/2022/06/01/669875.htm</a></li><li>AIR Worldwide corporate history and founding. Wikipedia. <a href="https://en.wikipedia.org/wiki/AIR_Worldwide">https://en.wikipedia.org/wiki/AIR_Worldwide</a></li><li>Karen Clark career profile and catastrophe modeling history. Triple-I Blog, March 2021. <a href="https://insuranceindustryblog.iii.org/womens-history-month-karen-clark-a-model-of-success/">https://insuranceindustryblog.iii.org/womens-history-month-karen-clark-a-model-of-success/</a></li><li>Karen Clark &amp; Company corporate history. <a href="https://www.karenclarkandco.com/company/about-kcc">https://www.karenclarkandco.com/company/about-kcc</a></li><li>Karen Clark, “A Formal Approach to Catastrophe Risk Assessment and Management,” <em>Proceedings of the Casualty Actuarial Society</em>, 1986.</li><li>Purdy House Institute Memorandum Series. Zenodo: <a href="https://doi.org/10.5281/zenodo.18007387">https://doi.org/10.5281/zenodo.18007387</a> | SSRN: <a href="https://papers.ssrn.com/sol3/cf_dev/AbsByAuth.cfm?per_id=7625798">https://papers.ssrn.com/sol3/cf_dev/AbsByAuth.cfm?per_id=7625798</a> | Academia.edu: <a href="https://independent.academia.edu/RyanJamesPurdy">https://independent.academia.edu/RyanJamesPurdy</a></li></ol><p><strong>Tags:</strong> History, Money, Culture, Psychology, Economics</p><p><strong>Bio:</strong> Ryan James Purdy is a leading AI Assurance Advisor and founder of the Purdy House Institute. He has spent 30 years in education across North America, Europe, and Asia. Connect on <a href="https://www.linkedin.com/in/purdyhouse/">LinkedIn</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=13f9a7710548" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>