Dan Hendrycks says numerous AI safety groups backed the MisAnthropic Statement in private but feared that signing could jeopardize relationships or future funding. Source
The statement’s website identifies Hendrycks as executive director of the Center for AI Safety (CAIS). In a November 18 quote-post of CAIS’s announcement, he wrote: “Anthropic has almost, but not entirely, captured the ecosystem meant to hold it accountable.” Source Source Source
His post doesn’t name the groups or describe an instance of Anthropic threatening or withdrawing funding. Source
The safety label is the target
“Anthropic warns that recursively self-improving AI could concentrate power and escape human control, yet races to build it first,” the statement says. Recursive self-improvement means AI getting better at improving itself. Source
The statement also accuses Anthropic of using safety branding to attract investors, employees and the public while weakening its safety pledges and lobbying for what the signatories call “self-serving regulation.” Source
This isn’t a request to polish the safety pitch. “Anthropic is not a safe AI company,” the statement declares. Source
The website lists Hugging Face co-founder Thomas Wolf, psychologist Steven Pinker, legal scholar Laurence Tribe and Humane Intelligence founder Rumman Chowdhury among its signatories. Those individual listings don’t establish institutional endorsements. Source
Comments