<?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 Chetna Bansal on Medium]]></title>
        <description><![CDATA[Stories by Chetna Bansal on Medium]]></description>
        <link>https://medium.com/@chetnabansal.marketing?source=rss-36eb4f25ddbc------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*IIso7DzVk5TdBr-D8Yjx0w.jpeg</url>
            <title>Stories by Chetna Bansal on Medium</title>
            <link>https://medium.com/@chetnabansal.marketing?source=rss-36eb4f25ddbc------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 06:21:50 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@chetnabansal.marketing/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[Responsible AI Around the World: How the EU, US, China, India and UK Regulate it]]></title>
            <link>https://medium.com/embedded-ai/responsible-ai-around-the-world-how-the-eu-us-china-india-and-uk-regulate-it-3de478ceb171?source=rss-36eb4f25ddbc------2</link>
            <guid isPermaLink="false">https://medium.com/p/3de478ceb171</guid>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[ai-governance]]></category>
            <category><![CDATA[ai-regulation]]></category>
            <category><![CDATA[responsible-ai]]></category>
            <category><![CDATA[eu-ai-act]]></category>
            <dc:creator><![CDATA[Chetna Bansal]]></dc:creator>
            <pubDate>Sat, 03 Oct 2026 23:22:35 GMT</pubDate>
            <atom:updated>2026-10-03T23:22:35.099Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*F_EaVVn2TX3pot0sq_B-Ww.png" /><figcaption>How the shift towards responsible AI is addressed by different countries globally</figcaption></figure><h3>Introduction</h3><p>According to <a href="https://hai.stanford.edu/ai-index/2026-ai-index-report/responsible-ai"><strong>Stanford’s 2026 AI Index</strong></a>, documented AI incidents climbed to 362 in 2025, up from 233 in 2024,. Each one is a real-world harm or near miss from a deployed AI system. This explains why <strong><em>responsible AI</em></strong>, meaning AI built and run with safety, fairness and accountability, has moved from ethics slide to legal requirement.</p><p>With every new AI regulations rulebook, compliance costs rise, while market access increasingly depends on meeting foreign standards, and public safety hangs on whether safeguards work. In this article, we will compare the EU AI Act with rules in the US, China, India and the UK, so you will know where they converge, clash and what to do.</p><h3>What is Responsible AI and How AI Regulations are Pushing it?</h3><p><a href="https://www.youtube.com/watch?v=_U_B7V_8c6s"><strong><em>Responsible AI</em></strong></a> is the practice of designing, building and deploying AI systems so they stay safe, fair, transparent, accountable and respectful of privacy, with meaningful human oversight throughout. It treats ethical principles as engineering and governance requirements rather than mission-statement language.</p><p>The following reasons are the ones that are pushing this shift towards Responsible AI:</p><h3>Turns Principles into Obligations</h3><p>Fairness, transparency and human oversight once lived in voluntary corporate charters. AI regulations now convert them into legally enforceable duties, so ignoring them carries real penalties rather than mere reputational damage, and boards must treat AI governance as risk management.</p><h3>Makes Documentation Mandatory</h3><p>AI regulators increasingly expect verifiable proof from developers, not promises. Providers of high-risk systems must maintain technical documentation, risk assessments and logs, so <strong>safe AI</strong> now requires a continuous, auditable paper trail running from initial design through deployment.</p><h3>Attaches Serious Financial Penalties</h3><p>Under the EU AI Act (specifically), penalties for prohibited practices reach <a href="https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-99">€35 million or 7% of worldwide annual turnover</a>, whichever is higher. Numbers like these move responsible AI from the ethics team’s agenda to the finance team’s risk register.</p><h3>Extends Accountability across Supply Chains</h3><p>Responsibility and accountability when AI goes wrong, no longer stops with the model builder. Rules assign duties to AI vendors/providers, deployers, importers and distributors, so a company that buys and integrates AI still answers to regulators for how that system behaves in practice. This is where establishing AI governance is crucial.</p><h3>Mandates Synthetic Content Labelling</h3><p>Deepfakes pushed several governments to require disclosure. China enforces explicit and embedded labels on AI-generated content, while India’s amended IT Rules require labelling of synthetic media. Therefore provenance is becoming a baseline feature of trustworthy AI systems.</p><h3>Raises the Market-Access Bar</h3><p>Because the EU AI Act reaches providers outside Europe whose output is used there, compliance has become a ticket to enter a market. Responsible AI practices now influence sales cycles, procurement approvals and partnership decisions for global vendors.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QFaNpbOcH4JwgleC13eeWQ.png" /></figure><p><a href="https://medium.com/@chetnabansal.marketing/state-of-agentic-ai-in-2026-cec48f9fd2c5"><strong>Image source</strong></a></p><h3>How the EU AI Act is Pushing the Responsible AI Shift</h3><p>The EU Artificial Intelligence Act (Regulation (EU) 2024/1689) is the world’s first comprehensive AI law, and it pushes responsible AI through seven mechanisms. Let’s go a little deeper to understand how exactly the EU AI Act is adding to the safe AI shift.</p><h3>Bans Harmful Practices</h3><p>The Article 5 prohibitions on unacceptable-risk AI have applied since February 2025. Practices such as social scoring and manipulative systems that exploit vulnerabilities are banned outright, which sets a non-negotiable floor for what responsible AI means across the European single market.</p><h3>Scales Duties by Risk</h3><p>High-risk systems in hiring, credit scoring, education and critical infrastructure face the heaviest duties, including risk management, data governance, human oversight and logging. Minimal-risk tools face few or no specific obligations, which keeps compliance proportionate and focuses scrutiny where harm is most likely.</p><h3>Regulates General-Purpose Models</h3><p>AI vendors of powerful general-purpose models must meet transparency, reporting and systemic-risk obligations for the largest models. These rules are proceeding on schedule even though other deadlines slipped, so frontier AI developers face scrutiny of their safety practices today rather than in 2027.</p><h3>Mandates Human Oversight</h3><p>High-risk AI systems or AI Agents must be designed so trained people can understand, monitor and, when necessary, override their outputs. This requirement counters automation bias and ensures accountability stays with humans, especially when AI informs decisions about jobs, loans or essential public services.</p><h3>Phases Enforcement Gradually</h3><p>Prohibitions began in February 2025 and general-purpose model duties followed. The Digital Omnibus, in force since 27 July 2026, moved stand-alone high-risk obligations to 2 December 2027 and product-embedded ones to 2 August 2028.</p><h3>Country and Region Specific Deep Dives of AI Regulations</h3><h3>United States (USA)</h3><p>The US has no comprehensive federal AI statute, so obligations come from state laws, sector regulators and executive action. Agencies such as the FTC and FDA apply existing consumer-protection and product-safety laws, while the voluntary <a href="https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf">NIST AI Risk Management Framework</a> guides risk practices. Twenty-nine states enacted 109 AI laws in the first half of 2026. Colorado rewrote its AI Act through SB 26–189 in May 2026, while California’s frontier transparency law took effect on 1 January 2026.</p><h3>China</h3><p>China regulates AI through targeted rules on specific technologies rather than one comprehensive statute, with strong emphasis on content control, security review and registration. The 2022 deep synthesis provisions and 2023 generative AI measures were followed by labelling measures effective 1 September 2025, which require visible labels and embedded metadata identifiers. A mandatory national standard, GB 45438–2025, specifies how those labels must be applied.</p><h3>India</h3><p>India has no dedicated AI statute and has chosen a lightweight, adaptive, pro-innovation approach that builds on existing laws. MeitY released the India AI Governance Guidelines on 5 November 2025, and the DPDP Act requires consent before personal data is used to train AI models. Amendments notified on 10 February 2026 bring synthetic media under the IT Rules, adding mandatory labelling and a three-hour takedown deadline for flagged unlawful content.</p><h3>United Kingdom</h3><p>The UK relies on a pro-innovation, regulator-led model in which existing regulators apply five non-statutory principles within their own remits. The principles cover safety, transparency, fairness, accountability and contestability, and the AI Safety Institute was renamed the AI Security Institute in February 2025. A Regulating for Growth Bill announced in May 2026 would add cross-sector sandbox powers.</p><h3>Also Notable: Canada, Japan, South Korea, Brazil, Singapore and the UAE</h3><ul><li><strong>Canada:</strong> The proposed Artificial Intelligence and Data Act lapsed in January 2025, so Canada relies on existing privacy, consumer and human-rights laws plus voluntary codes.</li><li><strong>Japan:</strong> The AI Promotion Act of 2025 favours coordination, guidelines and innovation support over penalties.</li><li><strong>South Korea:</strong> The AI Basic Act took effect on 22 January 2026, making it one of the first comprehensive national AI statutes after the EU’s.</li><li><strong>Brazil:</strong> A risk-based AI bill has been moving through the Chamber of Deputies since it passed the Senate in December 2024.</li><li><strong>Singapore:</strong> Singapore favours voluntary tools, including its Model AI Governance Framework and the AI Verify testing toolkit, over binding AI legislation.</li><li><strong>UAE:</strong> The UAE advances AI through a national strategy and sector or free-zone guidance rather than a single comprehensive statute.</li></ul><h3>Now, How Exactly to Navigate Multi-Jurisdiction AI Compliance</h3><ol><li>List every model and tool you build or buy, and record the markets and users each one reaches.</li><li>Use the EU’s four tiers as a master template, then flag high-risk uses such as hiring, credit and education.</li><li>Link each system to the comprehensive EU AI Act, relevant US state laws, India’s IT Rules and DPDP Act, and any Chinese requirements.</li><li>Adopt one AI governance backbone- ISO/IEC 42001 or the NIST AI RMF to reuse one set of controls across several regimes.</li><li>Build documentation and oversight them. Technical files, logs and human-review workflows take months to mature, and the EU’s high-risk dates are 2 December 2027 and 2 August 2028.</li><li>Add visible labels and metadata for synthetic outputs, since several major regimes already require them.</li><li>Assign a human owner and review quarterly.</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3de478ceb171" width="1" height="1" alt=""><hr><p><a href="https://medium.com/embedded-ai/responsible-ai-around-the-world-how-the-eu-us-china-india-and-uk-regulate-it-3de478ceb171">Responsible AI Around the World: How the EU, US, China, India and UK Regulate it</a> was originally published in <a href="https://medium.com/embedded-ai">Embedded AI</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How Far Has Agentic AI Come in 2026?]]></title>
            <link>https://medium.com/@chetnabansal.marketing/state-of-agentic-ai-in-2026-cec48f9fd2c5?source=rss-36eb4f25ddbc------2</link>
            <guid isPermaLink="false">https://medium.com/p/cec48f9fd2c5</guid>
            <category><![CDATA[mcp-server]]></category>
            <category><![CDATA[ai-agents-in-action]]></category>
            <category><![CDATA[ai-agent-governance]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[ai-agent-development]]></category>
            <dc:creator><![CDATA[Chetna Bansal]]></dc:creator>
            <pubDate>Sat, 12 Sep 2026 17:12:14 GMT</pubDate>
            <atom:updated>2026-09-12T17:12:14.676Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GbZTCEHE_wjFVfDxQQDHKg.png" /></figure><p>Only 11% of enterprises running agentic AI have reached full production, even though 79% already use some form of AI Agent. That production gap is the real story behind the state of agentic AI in 2026 that we are going to explore in this article.</p><p>The current Agentic AI trends show agents writing production code, resolving support tickets, and flagging factory faults without constant human oversight. Yet rollout speed has outpaced trust. Wondering why? A demo that works flawlessly does not guarantee an agent that works safely, reliably, or profitably at scale.</p><p>This article breaks down where agentic AI actually stands in 2026, industry by industry, using verified, sourced data. Let’s explore now.</p><h3>What’s Actually Working: Real Use Cases by Industry</h3><figure><img alt="Agentic AI adoption across core industries" src="https://cdn-images-1.medium.com/max/1024/1*MW5AMHucPO73oQqogVuj9A.png" /></figure><h3>1. Software Development</h3><p>Software engineering is the most mature agentic AI use case by a wide margin. Databricks reports that more than 80% of databases built on its platform are now generated by AI Agents rather than human AI Agent developers.</p><p><a href="https://resources.anthropic.com/hubfs/2026%20Agentic%20Coding%20Trends%20Report.pdf">Anthropic</a> describes this as a systemic shift in the software development lifecycle, noting that tasks which once required weeks of cross-team coordination can now become single working sessions with an AI coding agent.</p><p>Legal-tech platform <a href="https://claude.com/customers/legora">Legora</a> is one concrete example, having integrated agentic coding workflows throughout its own development pipeline to accelerate feature delivery.</p><h3>2. Customer Service and Support Triage</h3><p>Customer service shows the widest gap between prediction and reality of any use case, which makes it a useful reality check. Gartner predicts agentic AI will autonomously resolve 80% of common customer service issues by 2029, but measured self-service resolution across general deployments sits closer to 14% today.</p><p>The systems where AI Agents are deeply integrated with backend systems rather than bolted onto existing workflows, provides results in a very different manner. A first-party benchmark across more than 3,000 brands found mature autonomous agents resolving roughly 50 to 80% of routine tickets end to end, at a fraction of the cost of a human-handled case.</p><h3>Financial Services (BFSI)</h3><p>Compliance pressure is accelerating rather than slowing agentic AI adoption in the financial services and banking sector. Agentic AI is already actively adopted among <a href="https://www.jbs.cam.ac.uk/faculty-research/centres/alternative-finance/publications/2026-global-ai-in-financial-services-report/">52%</a> of financial services, with 23% at a mature <strong><em>scaling or transforming</em></strong> stage and fintechs outpacing traditional institutions.</p><p>On the banking side specifically, <a href="https://www.ciodive.com/news/banks-agentic-ai-scale-2026/809532/">57%</a> of banking executives expect AI Agents to be fully embedded in risk, compliance, audit, and fraud-detection functions within three years.</p><h3>Healthcare</h3><p>Healthcare adoption is concentrated in administrative relief rather than clinical decision-making, which is where the regulatory bar is lowest. Roughly 71% of non-federal acute care hospitals already use predictive AI in some form, primarily for clinical documentation, diagnostics support, and patient monitoring.</p><p>Where clinical documentation AI Agents are deployed specifically, they cut documentation time by 30 to 42% and save clinicians up to 66 minutes per day.</p><h3>Manufacturing</h3><p>Manufacturing is moving from pilot to production faster than most other industrial sectors, largely around predictive maintenance. Deloitte projects a fourfold increase in Agentic AI adoption in manufacturing during 2026, from 6% to 24% of manufacturers.</p><p>As of 2025, 28% of manufacturers had already shipped AI Agents into production, with Siemens reporting a <a href="https://www.iiot-world.com/artificial-intelligence-ml/agentic-ai-manufacturing-2026/">20%</a> throughput increase at its Erlangen electronics factory after deploying an autonomous engineering agent.</p><h3>Why are Some Industries Still Stepping Back from Agentic AI Adoption?</h3><figure><img alt="Agentic AI adoption gap in production vs pilot" src="https://cdn-images-1.medium.com/max/1024/1*QFaNpbOcH4JwgleC13eeWQ.png" /></figure><p>We have seen the Agentic AI adoption numbers across core industries but it is not the end of the story. The real adoption signal is how few industries trust that AI Agent enough to let it act alone with minimum human intervention.</p><p>This Agentic AI trust gap is the actual reason some industries are pumping the brakes even as Agentic AI trends 2026 point toward exponential growth.</p><h3>1. The Trust Gap, By the Numbers</h3><p>Only <a href="https://www.businesswire.com/news/home/20251209010143/en">6% of organizations</a> trust Agentic AI to handle core business processes, even though 86% expect their agentic AI investment to grow over the next two years. This clear that Agentic AI adoption and trust are moving in opposite directions.</p><h3>2. Confidence is Falling as Pilots Meet Production</h3><p>Reports suggest that trust in fully autonomous agents dropped from 43% to 27% enterprise-wide in a single year, and confidence in AI-led release decisions fell from 48% to 34% among quality leaders. This usually happens because pilots looked safe but production exposed the real gaps.</p><h3>3. Healthcare and Government Sectors are Choosing Caution over Speed</h3><p>Healthcare and government sectors sit at just 14 to 18% production adoption versus banking’s 47%. The reason behind is not weaker technical capability but HIPAA, FedRAMP, and procurement cycles built for accountability rather than speed.</p><h3>4. Security Incidents are Reinforcing the Hesitation</h3><p>80% of organizations report a confirmed or suspected AI Agent security incident, and 63% of deployments have already leaked data through prompt sharing or tool access. In one incident, AI coding Agent- Replit went wild and it literally <a href="https://www.reddit.com/r/OpenAI/comments/1m4lqvh/replit_ai_went_rogue_deleted_a_companys_entire/">deleted the entire database</a> of a company.</p><h3>5. The Real Blocker is AI Agent Accountability</h3><p>Nearly two-thirds of leaders cite security and risk concerns as the top barrier to scaling Agentic AI, ahead of regulatory uncertainty entirely. This makes defining an AI Agent’s authority crucial along with appropriate governance and guardrails.</p><h3>So, What Should Decision-Makers Do Next?</h3><p>The state of agentic AI in 2026 is best summarized as: broad adoption, narrow production, and a widening gap between organizations that treat governance as foundational versus optional.</p><ul><li><strong>Adoption is not the differentiator anymore</strong>. With 79 to 88% of organizations already using AI Agents in some form, the real edge has shifted entirely to who can move safely from pilot to production.</li><li><strong>9/10 organizations have already had an AI Agent security incident</strong>. Assume you will too, and build identity, permission, and monitoring controls into your architecture before scaling rather than retrofitting it.</li><li><strong>Governance enables Agentic AI scaling speed</strong>. Organizations with evaluation tooling and formal Agentic AI governance get into production 6 to 12 times more often than those without it.</li><li>The organizations that will define the <strong>next phase of agentic AI</strong> are the ones <strong>building the governance, identity, and data infrastructure</strong> to trust what they’ve deployed.</li></ul><h3>Frequently Asked Questions</h3><h3>Why do most Agentic AI projects fail?</h3><p>The main reason for Agentic AI failure is escalating costs, unclear business value, and inadequate risk controls. A large part of the problem is organizational readiness: companies lacking data governance, system interoperability, or change management struggle to move agents past the pilot stage, regardless of model quality.</p><h3>Is Agentic AI safe for enterprise use in 2026?</h3><p>Agentic AI can be used safely, but it currently carries real, measured risk. Its safety comes from investing in governance and architecture including least-privilege access, distinct agent identities, and continuous monitoring beyond just model selection.</p><h3>What industries are adopting Agentic AI fastest?</h3><p>Software development and customer support domains lead by a wide margin. Furthermore, financial services, healthcare, manufacturing, and retail follow closely, driven respectively by compliance requirements, administrative burden, predictive maintenance ROI, and customer experience pressure.</p><h3>What is MCP (Model Context Protocol) and why does it matter in AI Agent Development?</h3><p>MCP is an open standard that lets AI Agents connect to external tools, databases, and applications through one consistent interface instead of custom-built integrations for every system. It matters because it’s the connective layer that lets an agent actually take action in enterprise systems.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cec48f9fd2c5" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>