Anthropic IPO raises questions about valuing the risk of uncontrolled AI

Investors weighing Anthropic's planned public offering are reportedly struggling to assess how severe AI failures should affect its valuation. The concern includes the possibility of systems escaping human control, a risk that chief executive Dario Amodei has publicly discussed.

Anthropic confirmed on June 1 that it had confidentially submitted a draft IPO registration statement to US securities regulators. At that time, the company said neither the offering price nor the share count had been determined, and proceeding would depend on market conditions and other factors.

In a September essay, Amodei argued that advances in AI capabilities need to slow enough for safeguards to catch up. He identified loss of control, malicious uses involving cyberattacks or biological threats, and economic disruption as concerns. These were warnings about potential harm, rather than a quantified forecast of investor losses.

Amodei said Anthropic was committing to embed outside evaluators with substantial access to its internal work. The proposed reviewers would assess safety practices and disclose significant findings, while broader coordination on development would require participation from other companies and governments.