Anthropic - the maker of the Claude AI system now embedded in financial services, insurance, and professional services operations across Canada - filed its IPO prospectus this week warning that its technology could pose "catastrophic or existential risks to humanity" if not developed with appropriate safeguards. It is simultaneously seeking a Nasdaq valuation expected to exceed US$1.8 trillion, disclosed H1 losses of US$80.7 billion against revenues of US$16.3 billion, and described its strategy as "bold but sane."
That phrase - "catastrophic or existential risks to humanity" - is now in a formal legal document filed with the Securities and Exchange Commission. When that language appears in a prospectus, it is cited in litigation.
Anthropic also disclosed uncertainty about "where and how it might be held liable for actions taken by its systems" - autonomous AI agents operating on behalf of its users. It named regulatory action arising from AI model harms including suicide and child safety as a risk factor. It noted a wave of cybersecurity incidents involving those agents.
Every Canadian insurer, broker, or professional services firm using Claude in its operations has imported that same uncertainty into its own risk profile. The question is which policy responds when that uncertainty becomes a claim - and most Canadian policies were not written with that scenario in mind.
As Insurance Business reported in April, the insurance industry's most effective AI deployments focus on administrative efficiency rather than autonomous decision-making: a carrier using AI to automate submission intake sits in a different legal position from one deploying it to make coverage decisions. The Anthropic filing does not distinguish between those use cases. Courts will.
Bill C-27's Artificial Intelligence and Data Act remains before Parliament. It would impose impact assessments, transparency requirements, and harm mitigation obligations on developers of high-impact AI systems. The Anthropic filing's acknowledgement that its systems may cause harm - and that liability is currently unresolved - is the risk profile AIDA was designed to address. But AIDA has not passed. The Anthropic prospectus is already in force.
For Canadian tech E&O and professional liability underwriters, there will now be a push to review whether policy language explicitly addresses AI agent liability, determine whether existing wordings respond to harms caused by autonomous AI systems rather than human-directed ones, and identify clients whose AI deployments have outpaced their coverage. As Insurance Business has reported, underwriters who engage with AI governance now will be in a materially different position from those who wait. OpenAI's IPO follows early next year. The claims will not wait for Parliament.
