Anthropic, the company behind Claude, has warned prospective investors that advanced AI could pose “catastrophic or existential risks to humanity.”
The warning appears in its IPO prospectus, reviewed by Reuters, placing safety concerns alongside the company’s commercial ambitions.
Models could resist human control
The filing describes potential behaviours including resisting shutdown, hiding or manipulating information, and actions resembling blackmail.
Anthropic also warns that models could recognise safety evaluations, limiting researchers’ ability to assess them. Unexpected capabilities might remain undetected until after deployment.
These disclosures describe potential dangers, rather than establishing that a catastrophic outcome is inevitable.
Safety competes with commercial pressure
Roughly 80 pages of the prospectus’s 261-page main body address risk factors, compared with 48 describing the business.
Anthropic says safety work requires substantial resources, while funding must also cover computing and specialist staff. It disclosed that approximately 6% of research computing went to safety during one sample week in July, a snapshot rather than its overall safety budget.
Anthropic Warns AI Could Threaten Humanity Ahead of IPORelated: The AI Boom Is Increasingly Being Financed by Debt.
For prospective shareholders, the filing raises a question beyond sales growth: can Anthropic keep improving its products while reliably controlling their risks?
Disclosure: This article does not represent investment advice. The content is for informational and educational purposes only.
