@pactai_org founding member @ArmillaAI’s perspective as an insurer is one all organizations deploying AI should hear. Every AI liability policy they write is a judgment about whether a system will perform as promised, so the quality and consistency of independent evaluation matters as much to them as it does to the companies deploying AI.

"Insurance is how markets have always turned uncertainty into something they can act on, but that only works when there's credible evidence about how a system performs. Right now that evidence is inconsistent, and everyone in the chain pays for it: the deployer, the evaluator, the underwriter. PACT AI is building the shared infrastructure to fix that, and we want AI insurance to be one of the mechanisms that turns good assurance into real economic value for the companies that invest in it." -@_kramki, Founder & CEO of @ArmillaAI 🧵

Your customer's procurement team reads your certificate for one thing: does it name AI. That's where "AI-affirmative" and "AI not excluded" stop being synonyms. A silent policy doesn't mention AI. It might respond to an AI loss, but on its face, it can't satisfy a contract requirement for AI-specific coverage. An affirmative policy names AI as a covered exposure, with triggers written for AI failure modes: model errors, hallucinations, agent failures. AI is underwritable when you evaluate the AI. That is the policy we built. Contact us to learn more: armilla.ai/contact-us-ai-ins…

One of the things that makes AI risk fundamentally different from traditional technology risk is that AI systems change their behaviour over time. A software application does the same thing every time you run it. An AI model can drift. Its accuracy can degrade. Its outputs can shift as real-world data diverges from training data. This has profound implications for insurance. You can't write a static policy for a dynamic risk. The assessment you did at deployment isn't valid six months later. That's why Armilla's underwriting process includes ongoing technical evaluation. When we assess an AI model for insurance, we evaluated the system's architecture, monitoring capabilities, and resilience to drift. For enterprises: your customers need assurance that your AI works as intended, not just at launch, but over time. Insurance that understands this distinction is what separates purpose-built AI coverage from everything else. Case study: armilla.ai/resources/armilla…

AI governance is becoming insurable. Here's what that actually means: Companies with documented AI governance frameworks (inventories, risk assessments, monitoring) can now get better insurance terms. Insurance creates a market incentive for governance. Better governance = lower premiums = lower cost of AI deployment. This creates a virtuous cycle: insurance drives governance, governance drives trust, trust drives adoption. The governance-insurance nexus is the most under-appreciated force in responsible AI. Armilla AI sits at the centre of it. Want to learn more? Let's talk. armilla.ai/contact-us

The coverage gap explained. What your existing insurance policies DON'T cover when it comes to AI: CYBER: Covers data breaches, network intrusions. Doesn't cover AI hallucinations, output errors, model drift. D&O: Covers board decisions. Doesn't clearly cover autonomous AI decisions made without human oversight. E&O: Covers professional errors. Doesn't cover AI-generated IP infringement or content liability. PRODUCT LIABILITY: Covers physical defects. Doesn't cover software output failures or algorithmic harm. The result: gaps right where AI risks concentrate. This is exactly why Vanguard AI exists, combining Armilla's AI Liability Insurance with Chaucer's Cyber and E&O into one comprehensive solution. Learn more: armilla.ai/resources/chaucer…

What Boeing and the 2008 financial crisis teach us about AI oversight: Both disasters involved private entities trusted to evaluate risk independently. Both failed because independence was compromised and oversight was under-resourced. Credit rating agencies were supposed to evaluate financial risk objectively. They didn't. The 2008 crisis followed. In the Boeing 737-MAX case, the oversight process broke down because the institutions meant to provide accountability couldn't keep pace with the complexity. As we build private governance mechanisms for AI (including insurance), these failures are essential lessons. Independence matters. Funding matters. Accountability matters. At Armilla AI, underwriting independence isn't a nice-to-have. It's foundational. We're rigorously assessing AI deployments, not rubber-stamping them. That's the only way market-based governance works. Want to chat? armilla.ai/contact-us

Most company execs deploying AI can't produce a complete list of what AI systems they have in production. If you can't inventory it, you can't govern it. If you can't govern it, you can't insure it. Step one isn't a policy document. It's a spreadsheet. This is the first thing we hear from brokers evaluating AI liability coverage for their clients. The companies that can answer the inventory question get better outcomes. The rest are guessing at their exposure.

AI insurance is following the same trajectory as cyber insurance. Armilla's February 2026 white paper with Lockton Re documents the parallel in detail. Here's the playbook, stage by stage: Stage 1: New tech creates risks that existing policies don't cover. (We're here.) Stage 2: Early losses force market recognition of coverage gaps. (Happening now.) Stage 3: Specialist insurers build purpose-built products. (That's Armilla AI.) Stage 4: Regulatory pressure makes coverage standard. Stage 5: The market matures into a multi-billion dollar line. The white paper makes a critical distinction: AI risks aren't a subset of cyber risks. They're a distinct and broader category. AI systems can cause harm without any breach occurring. They can degrade, drift, and underperform in ways that cyber policies were never designed to address. That distinction is why purpose-built AI insurance matters. Cyber coverage is necessary. It's not sufficient. Cyber went from $600M to $15B+ in 15 years. AI insurance will compress that timeline. The companies that defined early cyber insurance defined the market for a generation. That's the opportunity we're building for. Read the Lockton Re + Armilla white paper: global.lockton.com/re/en/new…

Insurance will do more for AI safety than regulation alone. Here's why: Regulation is slow. Insurance prices risk in real time. Regulation sets floors. Insurance incentivizes best practices. Regulation reacts. Insurance anticipates. This isn't theoretical. It's exactly how insurance improved safety in automotive, aviation, healthcare, and cyber: faster and more effectively than regulation alone. AI insurance is the next guardrail. Unlike regulation, it's market-driven, adaptive, and already available. Read our blog on insurance as AI's next guardrail: armilla.ai/resources/the-nex…

Agentic AI is already sitting between your wallet and the internet. And almost nobody is talking about the liability implications. AI agents now browse the web, execute purchases, send emails, make financial decisions with minimal human oversight. When an AI agent makes a bad purchase, sends a harmful email, or executes a wrong transaction on your behalf, who's liable? The user? The AI developer? The platform? The company that deployed it? The legal frameworks aren't settled. But the deployments are already live. For AI SaaS providers and enterprises deploying agentic systems, this is an exposure that traditional insurance simply does not cover. Purpose-built AI liability insurance matters now more than ever. Armilla AI is building coverage that evolves with the technology. Let's talk. armilla.ai/contact-us

What if the answer to AI governance isn't more government regulation alone? Hadfield & Clark's "Regulatory Markets" paper proposes something different: instead of governments building AI expertise in-house (which has proven difficult across multiple sectors), create a competitive market of private AI regulators. These regulators would be licensed by governments, compete on governance quality, and be accountable for outcomes. Insurance fits naturally here. Insurers already evaluate risk, set standards, and create financial incentives for compliance. AI insurance is market-driven governance. The future of AI governance is probably not regulation OR markets. It's both, working together. Armilla AI sits at that intersection. TechUK explores similar themes: techuk.org/resource/ai-insur…

Armilla AI offers up to $25M in limits per policyholder, backed by Lloyd's of London. Dedicated AI liability coverage at enterprise scale. Our insurance capacity partner ecosystem includes Chaucer Re, Axis Capital, and Convex Insurance. This is what the market has been asking for. Available now for companies domiciled in the US and Canada. FFNews: ffnews.com/newsarticle/insur…

In 1982, Berliner published a framework for evaluating whether any risk is insurable. The Geneva Association applied it to generative AI. The results tell us exactly where AI insurance is headed: RANDOMNESS: AI losses aren't random. They can be systematic. This challenges traditional actuarial models but doesn't make AI uninsurable. It requires new modelling approaches. LOSS FREQUENCY: Data is sparse. Few historical claims exist. But mid-term frequency is expected to grow rapidly as deployment scales. INFORMATION ASYMMETRY: High. Insurers struggle to verify how companies manage AI risk. This is where governance and observability become critical for insurability. LEGAL PERMISSIBILITY: Evolving rapidly. AI-specific liability law is being written in real time, creating both risk and opportunity for insurers. Conclusion: AI is insurable, but only by insurers willing to build new frameworks. That's Armilla AI. Full report: genevaassociation.org/sites/…

The Geneva Association, one of the most respected research bodies in global insurance, recognized Armilla AI as a standalone AI insurance provider. Here's why that matters: @TheGenevaAssoc doesn't make casual endorsements. Their research shapes how the global insurance industry thinks about emerging risks. Being named in their Gen AI report alongside the analysis of AI insurability means the market recognizes that purpose-built AI insurance isn't a future concept. It's a present reality. The report applies Berliner's 1982 insurability framework to AI risks. The conclusion: AI is insurable, but requires new approaches. That's exactly what specialist insurers like Armilla bring. Not retrofitted cyber policies. Purpose-built AI liability coverage. Read the full report: genevaassociation.org/sites/…