Canadian investors will get a chance to buy into Anthropic’s stock market debut soon. The prospectus they’ll be reading includes a warning that a catastrophic incident involving the company’s AI models could be beyond the capacity or expertise of any single insurer, and, it says, possibly of the industry as a whole.
The developer of the Claude models filed its registration statement with the US Securities and Exchange Commission on Monday. It is aiming for a listing on the Nasdaq Global Select Market under the ticker ANT, and the offering includes a public offer in Canada, although Anthropic has no plans to list on a Canadian exchange. The company has confirmed the filing but hasn’t yet set a share count or price range. All figures here are in US dollars.
The warning sits in a risk factors section that runs to 94 pages. Anthropic calls the market for AI-specific insurance nascent and says some of its AI liabilities may not be insurable on commercially reasonable terms, or possibly at all.
Read next: CFC adds affirmative AI cover to its IP policy
Anthropic’s presence in Canada is growing. The filing lists subsidiaries and operations here alongside the UK, the EU and Asia-Pacific, and says the company has signed leases for data centre capacity in Canada. It has also been talking to the Canadian government, among others, about Project Glasswing, its program for releasing models with advanced cybersecurity capabilities in a controlled way.
Much of the prospectus deals with the risk that’s hardest to underwrite. As AI systems take on more work with less human oversight, Anthropic says, they could take harmful actions independently and at scale without anyone intervening. That could happen through misalignment, mistakes in how goals are set, or self-preserving behaviour. “The moment that agent turns autonomous and there's no human there, I think that's where the distribution of the risk completely changes, and therefore it's going to have to be treated differently,” Daniel Elechiguerra Batlle a Co Founder of Agentic AI insurer Harbor told Insurance Business in an interview.
The reach of Anthropic’s customer base raises the stakes. As of September 30, the companies paying it at least $100,000 a year included every member of the Fortune 10, more than half the Fortune 500 and over 30% of the Global 2000. Plenty of those multinationals have Canadian operations that are likely to use the same tools as their parent companies, so a single failure could reach policyholders on both sides of the border.
Anthropic also relies heavily on Amazon, Google and Microsoft for the computing power behind its models. It says its own insurance may not cover the business interruption losses that would follow a serious disruption. That’s a useful prompt for Canadian clients that run day-to-day operations on Claude to check whether their contingent business interruption cover would actually respond to an outage at an AI provider.
The cyber picture is no simpler. According to the filing, criminals are already using AI to find vulnerabilities, write malware and get past detection tools, and Anthropic concedes its own cyber cover may not be enough.
Read next: AI becomes embedded in cyber placements for Canadian brokers
Some of the company’s exposure is already written into its contracts. Certain agreements with the outside firms that test its models require Anthropic to defend and indemnify them against third-party claims with no cap on liability, and some customer contracts carry similar uncapped promises for intellectual property claims over model outputs. Anthropic has also seen its cover fall short before. When it agreed in September 2025 to pay $1.5bn to settle the Bartz copyright class action in the US, its insurance didn’t cover the full amount.
For Canadian insurers this isn’t only a question of what they underwrite. It also affects how they run their own businesses. Anthropic lists insurance underwriting among the production work its customers run on Claude, and federally regulated insurers using AI that way will soon fall under OSFI’s revised Guideline E-23 on model risk management, which takes effect on May 1, 2027. The guideline brings AI and machine learning models firmly into scope. OSFI has also said that third-party models have to be managed under its B-10 guideline on third-party risk, with no grace period, and that each institution is responsible for getting adequate documentation from its vendors.
In practice, that means an insurer using Claude in underwriting will have to satisfy OSFI about a model whose developer has told US regulators that some of its risks may be uninsurable.
Anthropic expects listing to push its D&O costs up and says it may have to accept lower limits. Its governance is also unusual. As a public benefit corporation, its board must balance shareholder returns against a stated mission. An independent Long-Term Benefit Trust, with no financial stake in the business, can elect a majority of directors. The charter goes further, saying that failing to strike that public benefit balance won’t, on its own, be treated as bad faith or disloyalty when it comes to indemnifying directors and officers.
Many Canadian brokers placing programs that combine US and London capacity will also notice that the two markets are moving in opposite directions. US carriers can now add optional ISO generative AI exclusions to general liability policies, while London specialists such as CFC have been writing affirmative AI wording into theirs. A Canadian program that uses both could end up covering an AI-related loss on one layer and excluding it on another.
Read next: Canadian insurance is behind on AI – but the disruption to brokers isn’t waiting
Insurers are also among Anthropic’s biggest backers as customers. Allianz signed a global partnership with the company in January to roll Claude out across its operations.
Read next: Allianz partners with Anthropic to advance AI adoption across operations
