Anthropic warns its AI could end humanity – and your clients are using it

Anthropic makes Claude, an AI system used by financial services, insurance, legal, and healthcare organizations around the world. On Monday, the company filed documents for its planned initial public offering (IPO). In these documents, it warned that its own technology could pose “catastrophic or existential risks to humanity” if it was not developed with proper safety measures. At the same time, Anthropic announced plans to list its shares on the Nasdaq stock exchange under the symbol ANT. Its expected value is more than SpaceX’s $1.8 trillion, which would make the listing the largest in stock market history.

The company reported losses of more than $80.7 billion in the first half of 2026, compared with revenue of $16.3 billion. It said it would try to be “bold but sane.”

Consider what this means: one of the world's best-known companies focused on AI safety has officially warned that its technology could pose catastrophic or existential risks to humanity. Its Claude system is already used in insurance risk assessments, claims processing, customer service, and compliance work across the US. The warning is now part of a formal document filed with the Securities and Exchange Commission (SEC), the US agency that oversees financial markets. Institutional investors, their lawyers, and lawyers representing people who bring lawsuits may refer to this document for many years to come.

The most important point for insurers

One key warning appears in the 90 pages describing the company's risks. Anthropic said it is not clear “where and how it might be held liable for actions taken by its systems.” The filing also says that Anthropic's AI agents, which can carry out tasks with some independence, have already been involved in a series of cybersecurity incidents. If these AI agents cause harm, it is not clear who would have to pay for it.

This uncertainty affects more than Anthropic. Every US insurer that uses Claude has brought the same uncertainty into its own risk profile. Every professional liability or technology errors and omissions (E&O) insurance policy covering a company that uses AI agents could face a claim under legal rules that do not yet exist.

In the same week, people who own Elvis Presley’s music sued Anthropic. They accused the company of using hundreds of his songs to train Claude without permission. This ongoing copyright case shows that questions about who is legally responsible for AI are already being argued in court. They are not just problems that might appear when AI technology becomes more advanced.

As IBA reported in June, Acrisure cut 2,250 jobs and said AI was a reason. This was an important moment because the industry openly recognized the effects of AI on jobs and business operations. Anthropic’s IPO filing is a similar turning point for legal responsibility and insurance risk. There is a big difference between experts saying AI could cause harm and a company formally telling the SEC that it could. The first is a possible risk discussed in theory; the second is an official risk that investors and insurers must consider when working out its financial cost. That difference has now disappeared.

What happens next

OpenAI announced plans for its own IPO shortly after Anthropic. Its stock market listing is expected early next year. What history tells us about AI and insurance is that the insurance industry tends to respond to new types of risk after problems appear, rather than preparing for them in advance. Any insurance underwriter who is not reviewing policy wording related to AI this week is already behind.