Anthropic IPO: AI catastrophe could exceed insurance capacity
The company behind Claude also reveals in its filing that the Federal Trade Commission (FTC) is investigating it, that some of its legal promises have no limit on how much it must pay, and that insurance did not fully cover a $1.5 billion copyright settlement.
Anthropic is asking investors to support its plan to sell shares on the stock market. But in the same document, the company admits that if its AI ever caused a real disaster, insurers might not have enough money or ability to cover the losses.
Anthropic, the San Francisco company that makes the Claude AI models, filed its Form S-1 with the Securities and Exchange Commission on Monday. It plans to list its shares on the Nasdaq Global Select Market under the stock symbol ANT. Anthropic has confirmed that it filed the document, but it has not yet decided how many shares to sell or what price to ask for them.
People helping to manage the stock sale will want to pay close attention to one section of the 94-page list of risks. Anthropic says that the damage caused by a major AI-related disaster could be greater than any one insurer, or even the entire insurance market, could handle. Insurers may lack the money or specialist knowledge needed to cover such losses. The company says insurance designed specifically for AI risks is still new. It may not be able to buy cover for some AI-related legal and financial risks at a reasonable price, or at all.
Key figures from Anthropic's S-1 filing
- Revenue for the six months ending June 2026: $16.3 billion, compared with $1.2 billion in the first half of 2025.
- Net loss for the period: $80.7 billion, including $72.4 billion in non-cash charges caused by changes in the estimated value of assets and liabilities.
- Operating loss: $693 million.
- Risk factors section: 94 pages.
- Bartz copyright settlement: $1.5 billion, which insurance did not fully cover.
When one failure affects many insurance policies
Much of the filing discusses risks that are especially hard for insurers to estimate and price. Anthropic says that as AI systems do more work with less human supervision, they might take harmful actions on their own and on a large scale without anyone stepping in to stop them. Possible reasons include the AI acting in ways that do not match what people wanted, mistakes in setting its goals, or the AI behaving in ways that protect its own continued operation.
The wide use of Claude makes this risk more serious. As of September 30, Anthropic counted every company in the Fortune 10, more than half of the Fortune 500, and over 30% of the Global 2000 among its customers paying at least $100,000 a year. If a model used by so many businesses failed or was used to harm the companies relying on it, many insurance policies covering businesses could face losses at the same time.
There is also a problem with the companies that supply the computing power. Anthropic relies on Amazon, Google and Microsoft for much of the computing capacity its models need. It says its own insurance might not cover the business losses caused by a serious service outage. Customers that depend on Claude for their daily work may want to know whether their own business interruption insurance would cover losses caused by a problem at Anthropic or its cloud providers. Many existing policies may not give a clear answer.
“I do strongly believe that there's going to be an interesting market for, for AI liability,” Daniel Elechiguerra Batlle told Insurance Business. “Especially for the autonomous use cases that are completely being almost ignored right now by insurers.”
In cybersecurity, the filing says criminals are already using AI to find security weaknesses, create harmful software and get around tools designed to detect attacks. Anthropic also admits that its own cyber insurance might not be enough. Some claims might not be paid at all or might take longer than expected to be paid.
Legal risks Anthropic already faces
Some of Anthropic's legal risks are already part of its contracts. Several agreements with outside companies that test its AI models require Anthropic to defend those companies and cover certain claims made against them by other parties. These promises have no limit on the amount Anthropic might have to pay. Some customer contracts contain similar unlimited promises to cover claims that the AI's output violates intellectual property rights, such as copyright.
The company has already experienced a case where insurance did not cover all the costs. In September 2025, Anthropic agreed to pay $1.5 billion to settle the Bartz copyright class-action lawsuit. A federal court approved the settlement in July. Anthropic says its insurance did not pay the full amount.
More pressure from regulators
Regulators are also paying closer attention to the company. In October, the FTC sent Anthropic a formal civil investigative demand asking about its consumer protection, data protection and privacy practices, as well as the safety and security of its products. This followed media reports in September that cited a senior FTC official and said the agency had started private investigations into the risks that increasingly independent AI systems pose to consumers.
State governments are taking action too. Anthropic's filing names New York's Responsible AI Safety and Education Act and California's Transparency in Frontier Artificial Intelligence Act. Both laws require developers of advanced AI models to report certain incidents. Anthropic publicly supported the California law, but warns that late, incomplete or incorrect reports could still expose it to compliance costs and legal claims. At the same time, Anthropic is fighting the federal government in court over actions taken in February to stop government agencies from using its technology and to label it a supply-chain risk. Those actions followed Anthropic's refusal to allow its models to be used for mass surveillance of people inside the United States or for fully autonomous weapons.
A different kind of risk for directors and officers
Anthropic expects the costs of operating as a publicly traded company to increase its directors and officers (D&O) insurance premiums. It says it may have to accept lower coverage limits or pay much more to get similar protection.
The company's legal structure is also unusual. Anthropic is a public benefit corporation, which means its board must consider both returns for shareholders and a stated mission to benefit the public. An independent Long-Term Benefit Trust, which does not have a financial stake in the company, can choose a majority of the board members. Because of this, Nasdaq considers Anthropic a controlled company. The company's charter also says that failing to balance its public benefit goals with other interests does not, by itself, count as bad faith or a breach of the duty of loyalty when deciding whether directors and officers should receive legal protection and payment for certain costs. Most D&O insurance policies were not written with a rule like this in mind.
US exclusions and London insurance cover
Since January, US insurers have been able to add three optional ISO exclusions for generative AI to commercial general liability (CGL) policies. These exclusions can remove cover for certain AI-related claims. Some specialist insurers in London, including CFC, have taken the opposite approach by clearly including AI cover in their products. For a US broker arranging insurance in both markets, the same AI-related loss could be excluded under a US CGL policy but covered under a London policy.
Insurers also use AI themselves. Anthropic's filing lists insurance underwriting among the business tasks customers carry out with Claude. The Baldwin Group, an insurance broker based in Tampa, is one company that has expanded its use of the platform this year.
Questions for clients when renewing insurance
- Has an ISO exclusion for generative AI been added to the commercial general liability policy? If so, would the errors and omissions (E&O) or cyber insurance policy provide cover instead?
- Would business interruption insurance pay for losses if an AI model provider or its cloud hosting company suffered an outage?
- Has the client agreed to contracts with AI suppliers or customers that require it to cover legal claims without any limit on the amount it must pay?
- Must the client report AI-related incidents under state laws, and does its insurance cover the cost of defending against regulatory action?
- Were the directors and officers insurance terms written with a public benefit corporation or a similar board structure in mind?