Edition 018: Risk
In the last editions, we considered the costs associated with the development of AI systems, not only the loss of autonomy and power that previous editions have described, but the more systemic costs, involving the land, water, energy, and labor taken from places and people without the power to refuse.
Now, we’ll turn to the pervasive risk posed by the use of AI, and how that is being noticed by the market—specifically, the insurance market. You may ask yourself, why does it matter what the insurance market thinks about AI? It matters because insurance is the institution that is required to quantify what it believes the risk is worth—how much the potential harm will cost. The insurance industry can price fires, floods, car accidents, product defects, and medical malpractice. But right now, it’s refusing to price generative artificial intelligence.
On January 1, 2026, Verisk’s Insurance Services Office (ISO) released three optional endorsements for commercial general liability policies.[1] ISO forms function as the standard templates for American insurance industry. The broadest of the three new endorsements, CG 40 47, excludes coverage for bodily injury, property damage, and personal and advertising injury arising out of generative artificial intelligence.[2] By excluding them, the endorsement means that insurance companies won’t pay for any claims in those areas if they are caused by generative artificial intelligence.
The ISO definition of “generative artificial intelligence” covers any “machine-based learning system or model that is trained on data with the ability to create content or responses, including but not limited to text, images, audio, video or code.”[3] This definition is expansive, sweeping in the customer service chatbot, the marketing content generator, the AI voice-assistant, the AI-assisted coding tool, and frankly every AI feature now embedded by default in most all enterprise software that businesses use without thinking. You can’t even email without using AI today. It’s embedded into the operating system on your phone, Microsoft 365, Google Workspace, Salesforce, medical records systems, revenue management platforms, and much more. Most companies never agreed to this. The technology was deployed by the companies selling it, integrated into the products people already depend on, and the risk now is being passed through to the people and companies who cannot opt out.
Yet already, state regulators have approved more than 80 percent of the AI exclusion filings insurance carriers have submitted.[4] And a number of large insurance companies have filed their own language, some of it even broader than the ISO baseline.[5] According to a Harvard Law School Forum on Corporate Governance, “[m]ost of these exclusions purport to be near absolute in scope, precluding coverage in full for any claim in any way related, directly or indirectly to the usage of any AI.”[6]
The same exclusions are now moving through directors and officers (D&O) and errors and omissions (E&O) insurance policies, which most companies rely on to protect their directors, officers, and board members and other professionals from claims related to the companies’ operations. The Securities and Exchange Commission has named AI a 2026 examination focus, which means exposure to claims is rising at the same time that insurance companies are tightening exclusions.[7]
As standard insurance companies exit, the exposure is moving into the excess and surplus markets, where Lloyd’s syndicates are offering AI-specific products at prices most businesses cannot pay.[8] But what is happening here isn’t really about pricing. It’s about the refusal to price based on an inability to assess the scope of the risk.
The insurance industry has looked at generative AI and thrown up its hands. The losses are too systemic to price. They can’t be reasonably capped. A single failure in a foundation model, like ChatGPT, could result in claims by ten thousand companies and people.
Insurance is the institution that, more than any other in the American economy, is supposed to quantify what it believes a harm is worth. When that institution declines, that is the assessment. The risk is uninsurable, which is the industry’s word for too much. The companies that built these systems will keep selling them. And risk will weigh on the people who cannot refuse them.
***
[1] “General Liability Endorsements - Assault Or Battery, Generative AI, Human Trafficking.”PropertyCasualty360, 6 Oct. 2025, www.propertycasualty360.com/fcs/2025/10/06/general-liability-endorsements---assault-or-battery-generative-ai-human-trafficking/. Accessed 21 May 2026.
[2] “AI Exclusions in Insurance Policies: Broad Language, Uncertain Impact.” Policyholder Pulse, 13 Apr. 2026, www.policyholderpulse.com/ai-exclusions-insurance-policies/. Accessed 21 May 2026.
[3] “CG 40 47 01 26 Exclusion - Generative Artificial Intelligence,” PropertyCasualty360, 26 Aug. 2025, https://www.propertycasualty360.com/fcs/2025/08/26/cg-40-47-01-26-exclusion---generative-artificial-intelligence/. Accessed 21 May 2026.
[4] “The AI Exclusion Wave and the $4.7 Billion Specialty Insurance Opportunity.” Insurance Intel, Apr. 2026, insuranceintel.substack.com/p/the-ai-exclusion-wave-and-the-47. Accessed 21 May 2026.
[5] Verisk CG 40 47: What the New AI Exclusions Mean for Your Commercial Clients." Gridex, 1 Mar. 2026, gridex.dev/blog/verisk-ai-exclusions/. Accessed 21 May 2026.
[6] “The Hidden C-Suite Risk Of AI Failures.” Harvard Law School Forum on Corporate Governance, 22 Sept. 2025, corpgov.law.harvard.edu/2025/09/22/the-hidden-c-suite-risk-of-ai-failures/. Accessed 21 May 2026.
[7] "Companies Seek to Reduce AI-Related D&O Liability." Business Insurance, 14 Apr. 2026, www.businessinsurance.com/companies-seek-to-reduce-ai-related-do-liability/.
[8] "Insurers Draw Battle Lines on AI: New Policies Cover Hallucinations While Others Exclude AI." AI:Productivity, 7 Mar. 2026, aiproductivity.ai/news/ai-liability-insurance-coverage-exclusions-2026/. Accessed 21 May 2026.
