Anthropic IPO filing spends 69,800 words on investor risk, little on public costs

The modern prospectus, the commentary notes, evolved as a correction to railway-age manias, built to tell investors what could go wrong for them. Anthropic's filing performs that task with unusual zeal, cataloguing catastrophic risk, job displacement, rogue agents and industry power concentration as potential costs to Anthropic itself—even warning that its own public statements about AI risks "may be used against us in litigation."

The author, in a research collaboration with Oxfam, has assembled a Register of AI Risks & Remedies with 103 entries drawn from investigations, court proceedings, media and technical reports. That Register tracks costs already borne by third parties, often without consent or return. The commentary cites examples including a Dutch person mistakenly flagged by a fraud algorithm, a Polish doctor whose skills eroded through over-reliance on an algorithm, and a Kenyan content moderator who labeled toxic content.

The piece also points to a pattern of settlements—Detroit's $300,000 payment to Robert Williams after a flawed facial recognition match, SafeRent's $2.275 million settlement, and Anthropic's own $1.5 billion settlement over pirated books—and notes that about four out of five of the 103 documented harms were addressed by existing law rather than new AI statutes. It adds that 30 states enacted 115 AI-related laws this year, a nearly 58 percent increase, and that eight state attorneys-general are suing OpenAI, the FTC is investigating frontier labs' safety lapses, the EU is launching AI Act enforcement, and the Financial Stability Board is urging banks to strengthen AI governance.