Insurability of AI advancements

"This paper focuses on developing a conceptual blueprint for AI insurance that addresses unintended outcomes resulting directly from an AI system's normal operation, where outputs fall within the declared scope but diverge from intended behaviour. Such failures are already silently embedded in existing insurance portfolios, neither affirmatively covered nor excluded, and thus remain unpriced and unmanaged. We argue that dedicated AI insurance is necessary to quantify, price, and transfer these risks, while simultaneously embedding market-based incentives for safer and more secure AI deployment. The paper makes four contributions. First, we identify the core underwriting challenges, including the lack of historical loss data, the dynamic nature of model behaviour, and the systemic potential for correlated failures, and propose mechanisms for risk transfer and pricing, such as parametric triggers, usage-based coverage, and bonus-malus schemes. Second, we examine market structures that may shape the development of AI insurance and highlight technical enablers that support the quantification and pricing of AI risk. Third, we examine the interplay between insurance, AI model risk management, and assurance. We argue that without insurance, assurance services risk becoming box-ticking exercises, whereas underwriters, who directly bear the cost of claims, have strong incentives to demand rigorous testing, monitoring, and validation. In this way, insurers can act as guardians of effective AI governance, shaping standards for risk management and incentivising trustworthy deployment. Finally, we relate AI insurance to adjacent coverage lines, such as cyber and technology errors and omissions." Lukasz Szpruch Agni Orfanoudaki Carsten Maple Matthew Wicker Yoshua Bengio Kwok Yan Lam Marcin Detyniecki AXA

🚨 New Paper Alert 🚨 I’m excited to share a new article co-authored with Josephine Wolff The Limits of Regulating AI Safety Through Liability & Insurance. Read Here: https://lnkd.in/gjbdpU9s. This paper is motivated by the fact that many policymakers and scholars argue that if AI firms are held liable for harms, insurers will step in—pricing risk and incentivizing safer practices. The theory is elegant. The reality is messier. Our core argument: Insurers are unlikely to promote meaningful AI safety, just as they largely failed to do so in cybersecurity. Why? Cyber insurers were once hailed as “private regulators.” Two decades later, they still struggle to price risk in ways that improve security. AI presents even steeper challenges: scarce data, rapidly evolving systems, technical complexity, and risks embedded across business operations. Without strong ex ante regulation, liability insurance will at best pool losses after the fact—and at worst distort incentives, encourage secrecy, and create false confidence. This doesn’t mean insurers have no role. They can help firms manage narrow risks (like performance guarantees) and absorb liability shocks. But expecting liability and insurance to regulate AI safety is a mistake. ⚖️ Bottom line: Effective AI safety governance requires proactive regulatory frameworks, not reliance on liability insurance markets. Would love to hear reactions from those working on AI regulation, insurance, or liability.

AI Liability Insurance Coverage We have all seen issues involving AI mistakes, which raise issues of the liability arising out of them and whether insurance coverage is available for that liability. Apparently this type of coverage is now on the market. “The new product addresses the evolving risks of mechanical underperformance in AI systems and models, along with their associated liabilities.” “Coverage includes hallucinations (false or misleading outputs), model drift (performance degradation over time), mechanical failures, and other deviations from expected AI behaviour. It also provides legal defence and liability protection for claims arising from such underperformance.” Having seen so many examples of the use of AI by lawyers and the citation to nonexistent cases that resulted, I wonder if this coverage should be added to all legal malpractice policies? I also see the use of AI as a significant risk of breaching the intellectual property rights of others as it seems that AI takes text and graphics from anywhere it can find it on the internet. What AI related liability are you seeing - or anticipating seeing in the future? Should this be an add on or free-standing coverage for all liability policies? #LauraHasItCovered

Is your agent insurable? After being inspired by Joshua Saxe's keynote at the Offensive AI Con on the strategic imperative to drive our agents towards "meaningful autonomy," I've been thinking about what it will really take to get there. With the continued rapid advances in AI, the biggest hurdle to full autonomy won’t be a technical limitation but trust in our agents. The reality is that the old legal safeguards are failing in an era of “nuclear verdicts,” and a powerful demo is no longer enough to get an enterprise deal done when millions of dollars might be at stake if something goes wrong. This is why a new category of AI Underwriters is emerging, led by pioneers like Rune Kvist and Rajiv Dattani of the Artificial Intelligence Underwriting Company. Their work on the AIUC-1 standard is creating the "SOC 2 for AI" that the market is demanding. To put my thoughts together, I wrote a playbook on the controls required to build an agent that is not just highly capable, but legally defensible and commercially insurable. It's built on three pillars: → The Immutable Ledger: For forensic proof. → The Control Plane: For real-time safety. → Simulation: For actuarial evidence. The future of agents is both autonomy and accountability. Link in the comments, let me know what you think!