Tech Policy Advocacy

⚖️ Who’s Responsible When an AI Agent Makes a Bad Decision? A cancer diagnosis delayed. A pedestrian killed by a self-driving car. A loan denied due to biased data. The consequences are real. But the accountability? Still dangerously unclear. And in my opinion, the uncomfortable truth is that we’ve built autonomous systems… but forgot to build autonomous accountability. Over the past few months, I’ve been thinking deeply about this — especially as AI agents become more embedded in critical decisions across industries. 📚 One paper that really struck me recently is by Filippo Santoni de Sio and his colleagues. According to them, we’re facing a “Responsibility Gap” — not just one, but four intertwined gaps: ➤ Culpability ➤ Moral accountability ➤ Public accountability ➤ Active responsibility And these aren’t just tech issues — they’re legal, organizational, and societal. So what should we do? According to me, we’re falling into 3 common traps ❌ Fatalism: “This can’t be fixed.” Deflationism: “It’s not that serious.” Solutionism: “Tech and laws will solve it.” 👨⚖️ For business leaders, here’s why this matters: If AI is involved in your operations: ✅ You might be legally responsible — even if AI is “just assisting” ✅ Customers will hold you accountable ✅ Regulators are catching up — The EU AI Act shifts the burden of proof toward developers and deployers. 🛠️ In my opinion, we urgently need to explore emerging ideas to bridge the gap by adopting a few principles: 🔸 Computational Reflective Equilibrium – Accountability based on actual control 🔸 Presumption of Causality – Easier paths for victims to seek justice 🔸 Hybrid Governance – Audits, explainability, and regulation in one framework Here’s the big question for you: When AI causes harm… who should be held accountable? The CEO? The engineer? The algorithm? Or is it time to redefine liability in the age of autonomy? I’d love to hear what you think. #AI #Leadership #ResponsibleAI #Governance #IrreplaceableAI #AIRegulation #MeaningfulHumanControl #FutureOfWork

An AI policy is not AI governance. What matters is moving from policies to processes to daily practices. This new paper from the Vector Institute details a tactical, structured approach to AI Governance. Based on this paper, these 6 questions can help assess whether AI is moving from policy to practice: 1.    Do we know where AI is currently being used across the organisation – including Shadow AI? 2.    What governance processes are in place to evaluate the risks of new AI systems or use cases? 3.    Who is accountable for AI oversight – and are they appropriately supported with resources, authority, visibility, etc? 4.    How do we monitor the performance, safety and ethical use of AI systems post deployment? 5.    Do staff know what’s permitted and prohibited when using AI tools? Are they supported with training and guidance? 6.    Are we capturing and retaining the right records (e.g. inputs, outputs, design decisions) to create a defensible audit trail?

I helped build Facebook. I watched it become a machine for addicting people instead of connecting them. Because addiction was more profitable. Every social media company ran on the same logic: if we don't do it, someone else will. I haven't posted here in a long time. I'm back because that same logic is now driving AI — and the stakes are incomparably higher. Sam Altman, Dario Amodei, Demis Hassabis, Elon Musk, and Mark Zuckerberg all face the same trap: if I don't do it, someone else will. They're right. That's the problem. No single company can exit the race alone. Last week, the White House proposed a familiar answer: shield the industry from liability and let the companies sort it out. There's a better answer: put the public in charge of AI. Not politicians. Not regulators captured by the industry they're supposed to oversee. Everyday people — through citizens' assemblies, a model that's been working for thousands of years. They're already shaping AI policy in Taiwan, the UK, and Belgium. 66% of Americans support citizen panels helping set AI rules. That number holds across Trump voters, Biden voters, and swing voters — in a country that can barely agree on anything. Citizens don't write the code. They decide what the code should be for. With technical experts accountable to them for implementation. We built social media without giving the public a real say. We don't have to make that mistake again with AI. The piece is in Fortune — not where I expected my first op-ed on democratic AI governance to land. See comments for links.

The expansion of robots and automation is poised to significantly transform the job market and has complex implications for inequality. What do you think? Impact on Jobs: 1. Job Displacement: Robots and automation are likely to replace repetitive, manual, and routine jobs (e.g., manufacturing, logistics, and data entry). Some middle-skill jobs may also be at risk as automation technologies become more sophisticated. 2. Job Creation: New roles will emerge in robotics maintenance, programming, AI development, and other tech-focused fields. Demand for human-centric jobs, such as healthcare, education, and creative industries, may increase as these areas are harder to automate. 3. Job Evolution: Many jobs will change in scope, requiring workers to collaborate with robots or leverage automation tools for productivity. Impact on Inequality: 1. Widening Skill Gap: Workers with higher education and tech-savvy skills are more likely to benefit, while those in low-skill jobs may struggle to adapt. This divergence could exacerbate income inequality if reskilling programs are not widespread. 2. Geographic Disparities: Advanced economies with resources to invest in automation could benefit more than developing countries, increasing global inequality. 3. Ownership of Technology: Concentration of robot and AI ownership among corporations and wealthy individuals might widen wealth disparities unless equitable policies (e.g., profit sharing, taxes) are implemented. Mitigating Inequality: 1. Education and Reskilling: Governments and companies need to invest in upskilling and reskilling workers to prepare them for the jobs of the future. 2. Universal Basic Income (UBI): UBI or similar safety nets could help address income gaps caused by job displacement. 3. Fair Policies: Regulations around labor, taxation, and profit sharing could ensure that the economic benefits of automation are distributed more equitably. 4. Support for Vulnerable Sectors: Strengthening social welfare systems and providing targeted support for industries and workers most at risk. Video: @discover_our_planet_ #Innovation #Technology #Inequality

Over the last two weeks, both the U.S. Government and Anthropic took significant actions that demonstrated their power to control access to AI by restricting what others can do with frontier models. This has been one of those moments that, once seen, will be hard to unsee, and it is significantly accelerating many businesses’ and nation states’ efforts to ensure reliable access to AI that no one else can terminate. Anthropic first released Claude Fable 5, a version of its Mythos model with additional guardrails, including some restrictions that seem justified on safety grounds (such as limitations on applying it to hacking, bioweapons, and so forth). However, it also restricted developers’ ability to use it to build competing LLM technology. This move was concerning, given that the whole AI community, including Anthropic, has benefitted tremendously from open research — indeed, the AI revolution was kicked off by my former team (Google Brain) freely publishing the Transformers paper! Imagine if Microsoft’s terms of use barred anyone from using their tools to build competitive software, or if Google barred using it to search for information to work on competing search engines. Anthropic’s argument that it was unsafe for others to be able to make advances in AI also rang hollow. Initially, Anthropic silently degraded Fable 5’s performance for users detected to be working on LLM research through invisible interventions that weakened the model’s outputs without notifying the user. After significant backlash, it walked back this decision and decided to be transparent when it did this, but it still refuses to use its latest capabilities to help AI researchers. This move represents a raw demonstration of power by Anthropic. It has used “safety” arguments to hinder potential competitors. Platforms succeed when they are viewed as stable, reliable partners that one can build on. The sudden rule changes by Anthropic (including a mandatory 30 day data retention policy for Fable usage) have made developers wonder about the stability of building on any one proprietary LLM provider, not just Anthropic. The U.S. Government then shortly followed with an even greater demonstration of power. It used the Commerce Department’s authority to regulate technologies that may be national security threats to restrict exports of Mythos and Fable, requiring a license for use by any foreign national, whether inside or outside of the U.S. This led Anthropic to disable access to Fable to all users worldwide. Sam Altman pointed out, referring to Anthropic, “It is clearly incredible marketing to say, ‘We have built a bomb, we are about to drop it on your head. We will sell you a bomb shelter for $100 million.’” But when one uses this type of fear-based marketing, it increases the odds that the U.S. Government will agree with you and slap export controls on the bomb you say you have built. [Truncated for length. Full text: https://lnkd.in/gph-rGtw ]

🚨 BREAKING: China's new law on AI anthropomorphism has been officially enacted, and it is the world's STRICTEST law on the topic: As I wrote earlier this year, to my knowledge, no AI law anywhere in the world regulates anthropomorphic AI systems with this level of detail, strictness, and concern for context-specific vulnerabilities and potential risks. Earlier in January, I wrote an article about the law's first draft (link below). The approved version is even more comprehensive, covering liability-related risks as well. Article 10, for example, establishes that providers of anthropomorphic AI must fulfill their security responsibilities throughout the service lifecycle and sets out detailed obligations for each phase of AI development and deployment. Regarding children specifically, among the prohibited anthropomorphic AI practices is generating content for minors that causes them to imitate unsafe behaviors, induces extreme emotions, or leads them to develop bad habits, which may affect their physical and mental health. Despite being a serious topic (which has led to numerous cases of suicide and mental health harm), most countries do NOT regulate AI anthropomorphism comprehensively. An important reason for that is that peer-reviewed studies about AI-powered emotional manipulation and mental health harm only became available recently (as only in the past years have millions of people started to engage in these types of relationships). China's new law is worth taking a look at, and hopefully, other countries, states, and regions will soon follow suit with their own protections against AI anthropomorphism. 👉 Lastly, if you are interested in China's AI policy and regulation, besides joining my newsletter's 93,200+ subscribers, I invite you to join my new Masterclass on the topic (only on June 1st). Links below.

#Diversity in high-tech fields remains critically low. The Equal Employment Opportunity Commission (EEOC) recently reported that #Black and #Latino professionals are underrepresented in high-tech roles, especially in leadership. These numbers highlight ongoing structural barriers in hiring, promotion and retention. This gap is a missed opportunity to tap into a wealth of diverse talent and perspectives essential to the future of tech. However, addressing and thoroughly fixing these challenges will require time, consistent effort and a long-term commitment to systemic change. Companies can support the progression of representation in tech by investing in training, mentorship and internship opportunities that open doors for people who were historically shut out. Programs like internXL, a platform that is committed to increasing diversity and inclusion in the internship hiring process for top companies, are making a significant impact. Similarly, the expansion of STEM education at institutions like Cornell University is helping to connect talented young people from underrepresented communities with opportunities for high-tech careers. When we work together to remove these barriers, we’re fostering a more inclusive workforce and strengthening innovation, problem-solving and leadership in the industry. Let’s build a tech future that reflects the diversity of our society. https://bit.ly/3UNtOCh

Which country has the best government–startup relationship in the world? It’s a surprisingly rich question. And one with no single answer. Last month, the UK government appointed Alexandra Depledge, MBE as its first Entrepreneurship Adviser. Her task: tackling the key barriers faced by startups scaling in the UK - no small feat. Other countries have taken different directions. 🇮🇱 Israel: The Yozma model was decades ahead of its time, producing the world’s highest startup density per capita. It combined government risk-sharing with private VC through programs like Yozma, which offered matching funds and favourable buyouts. It helped create Waze, Mobileye and many NASDAQ-listed firms. Much of this was backed by Israel’s Office of the Chief Scientist (now the Israel Innovation Authority), a central force in early-stage tech funding and public-private innovation bridges. 🇪🇪 Estonia: e-Residency turned a small country into a digital powerhouse. Entrepreneurs can set up EU businesses remotely — attracting 120,000+ founders and €67m+ in tax revenue. 🇸🇬 Singapore: The most systematic approach. StartupSG grants and equity (with public/private funds), tax support, and structured business services. It’s the full package. 🇨🇱 Chile: Pioneered the government accelerator model, offering equity-free funding. It’s helped launch 1,800 international startups and build a talent pipeline into South America. 🇨🇦 Canada: Immigration, immigration, immigration. Entrepreneurs securing backing from designated investors can qualify for permanent residency. 🇦🇪 Dubai: Appointed the world’s first Minister of AI and launched innovation-friendly zones like DIFC and Dubai Future Foundation. Policies focus on frontier tech, digital commerce, and global talent. 🇺🇸 USA: Still the gold standard for scale and ambition. While lacking a central startup policy, R&D funding, DARPA, SBIR, and visas like the O-1 create a strong base. Crucially, its risk culture and VC depth do much of the heavy lifting. And then there are the UK and France... one with a new Treasury adviser, the other with unofficial founder back-channels (Xavier Niel and others DMing President Macron). So what works best? Successful models typically: ✔ Share risk (rather than grant cash) ✔ Provide regulatory clarity ✔ Build ecosystems, not just startups ✔ Attract international talent (and support local champions) ✔ Leverage national strengths (digital ID, military tech, tax regimes…) What doesn’t work? Overfunded but underambitious granterpreneurs relying on government rather than markets. Bureaucracy. Pilot programs that never scale. Would love your views. Which countries do this best? And what can the UK learn from them?

AI security/securing the use of AI is going to kill me. I use Claude Code almost daily. It's a problem.... Here's what I have to change AGAIN this week. Security researcher Ari Marzuk disclosed 30+ vulnerabilities across AI coding tools. Cursor. GitHub Copilot. Windsurf. Claude Code. All of them. He called it IDEsaster. The attack chain includes prompt injection, hijacking LLM context, and auto-approved tool calls executing without permission. Then, legitimate IDE features are weaponized for data exfiltration and RCE. Your .env files. Your API keys. Your source code. Accessible through features you thought were safe. Most studies I read claim that around 85% of developers now use AI coding tools daily. Most have no idea their IDE treats its own features as inherently trusted. 𝗦𝗼... 𝗮𝗳𝘁𝗲𝗿 𝗿𝗲𝘃𝗶𝗲𝘄𝗶𝗻𝗴 𝗔𝗿𝗶'𝘀 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵, 𝗵𝗲𝗿𝗲'𝘀 𝗜 𝘄𝗶𝗹𝗹 𝗯𝗲 𝗱𝗼𝗶𝗻𝗴... Be warned: All this is SO much easier said than done! Audit every MCP server connection. Checked for tool poisoning vectors where legitimate tools might parse attacker-controlled input from GitHub PRs or web content. Removed servers I couldn't verify. Disabled auto-approve for file writes. The attack chains weaponize configuration files and project instructions like .claude/settings.json and CLAUDE.md. One malicious write to these files can alter agent behavior or achieve code execution without additional user interaction. Move all credentials to a secrets manager. No .gitignored .env files in agent-accessible directories. API keys live in 1Password CLI. Environment variables inject at runtime through a wrapper script the LLM never sees. Start running Claude Code in isolated containers. Mounted volumes limited to specific project directories. No access to ~/.ssh, ~/.aws, or ~/.config. If the agent gets compromised, blast radius stays contained. Enable all security warnings. Claude Code added explicit warnings for JSON schema exfiltration and settings file modifications. These exist because Anthropic knows the attack surface. Add pre-commit hooks for hidden characters. Prompt injections hide in pasted URLs, READMEs, and file names using invisible Unicode. Flag non-ASCII characters in any file the agent might ingest. The fix isn't to stop using AI coding tools. The fix is to stop trusting them implicitly. What controls do you have for AI tools with write access to your codebase? 👉 Follow for more AI and cybersecurity insights with the occasional rant #AISecurity #DevSecOps

Microsoft: No More Data Center NDAs In a move to unwind years of secrecy and codenames, Microsoft said it will cease using non-disclosure agreements (NDAs) as a tool in its data center site selection and development. “We’ve made the decision that being transparent with the communities where we operate or seek to operate is paramount,” the company said. “This shift is about strengthening public trust, enabling better dialogue, and ensuring that our growth is matched by meaningful engagement.” NDAs have always been a trust-killer for community engagement, leaving residents with the impression that developers and local officials have something to hide. This issue has come up repeatedly as communities across the country mobilize to oppose large projects, and yet the NDAs have persisted. As we’ve noted regularly, local resistance has become a huge business issue for hyperscalers who need AI capacity ASAP. This has clearly gotten Microsoft’s attention, and this announcement is an important step in trying to rebuild trust with the communities where data centers will operate. https://lnkd.in/ewcF-6QA