A coalition convened by the Center for AI Safety is challenging Anthropic’s portrayal of itself as a safety-focused AI company ahead of its planned initial public offering. Its MisAnthropic Statement argues that the company’s public warnings about the dangers of recursively self-improving AI conflict with its pursuit of that capability as a business advantage. The signatories say this combination lets Anthropic attract safety-minded employees, investors, and public support while seeking to automate AI research and build influence. They come from varied fields and viewpoints, including AI safety, technology, law, academia, and open-source advocacy; the coalition says their shared concern is a company presenting its growing power as serving the greater good.
The statement points to three main grounds for its criticism: Anthropic’s pursuit of recursive self-improvement despite warning that it could concentrate power or evade human control; changes to its Responsible Scaling Policy, including dropping prior unilateral commitments to pause; and lobbying it characterizes as self-serving. It says the company sought to weaken California’s SB 1047, the RAISE Act, and the AEGIS Act, and advocated privately for federal pre-emption of state AI laws. CAIS executive director Dan Hendrycks argues that weakening safety commitments while seeking vast investment to build self-improving AI undermines the company’s claim to be safe. The release presents these points as the coalition’s case against Anthropic’s safety branding, not as a neutral assessment of the company.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.