The cover you assume you have is being rewritten
For years AI risk sat inside ordinary business policies by default. There was no mention of artificial intelligence in the wording, so a loss involving it was simply covered like any other - a state the market now calls silent AI. That implicit cover is reaching its end. In January 2026 the standard insurance market introduced a generative AI exclusion for commercial general liability policies, removing cover for bodily injury, property damage and personal or advertising injury that arises out of or is attributable to generative AI.
This is not a fringe product change. It is a standard endorsement that flows through the policies most companies hold without anyone reading the renewal line by line. The gap it creates is quiet by design - the premium looks the same, the policy number looks the same, and the carve-out only becomes visible when a claim is filed and denied. A company can be running AI agents in production while the insurance behind those agents has stopped covering what they do.
Named carriers are moving, and it is spreading into D&O
The shift is being driven by large insurers, not obscure ones. In November 2025 AIG, Great American and W.R. Berkley filed requests with US regulators for permission to exclude liabilities tied to AI tools such as chatbots and agents. Across the market, carriers are declining to write cover for AI-generated outputs in cyber and errors and omissions lines, raising rates to offset the risk, and in some cases declining to quote AI exposure at all.
The exclusions are not staying inside one policy type. Several insurers have introduced broad absolute AI exclusions in management lines, purporting to remove cover for claims arising out of any use, development or deployment of artificial intelligence - which reaches into directors and officers, employment practices and fiduciary cover. For an owner that means the exposure is no longer just an operational one; it can land on the people who sit on the board.
What to check before you deploy another agent
This is a development to act on, not advice on what to buy. The practical step is to read the renewal, not the summary. Pull your current general liability, cyber, errors and omissions and directors and officers wordings and look specifically for any AI or generative AI exclusion, and for how each policy defines AI. Then map a realistic AI failure - an agent that gives a customer wrong guidance, leaks data or makes a faulty automated decision - against those policies and find which line is meant to respond. Where the answer is none, you have a coverage gap, and you have found it before a claimant did.
Two things compound the risk. Insurers can treat undisclosed AI use as grounds to deny a claim later, so what you have told your carrier about your AI deployment matters as much as the wording itself. And because the absolute exclusions reach into board-level cover, this belongs in front of your broker and your legal counsel, not buried in a renewal folder. The owners who check now are the ones who will not learn the wording the hard way.
Read next: Your AI Bill No Longer Has a Ceiling · Match the Model Tier to the Job