Hot AI News

Pinker, Tribe and Tegmark sign a letter saying Anthropic is not a safe AI company

The Center for AI Safety and a broad group of signers say Anthropic is not a safe AI company, days before its IPO. The claims are theirs; Anthropic's reply is not in the documents.

Pinker, Tribe and Tegmark sign a letter saying Anthropic is not a safe AI company
Image: The MisAnthropic Statement

The Center for AI Safety (CAIS) and a broad group of scientists, legal scholars and tech figures have signed “The MisAnthropic Statement”, an open letter that says Anthropic is not a safe AI company. CAIS published it on November 18, ahead of Anthropic’s IPO, and Bloomberg covered it a day later.

Everything below is the signers’ position, not established fact. Neither the CAIS page nor the statement site includes a response from Anthropic.

What the signers say about Anthropic

According to the CAIS release and the statement site, the letter makes four claims. First, Anthropic warns that self-improving AI could concentrate power and slip out of human control, yet races to build it. Second, it uses safety branding to win over investors, employees and the public. Third, it weakened its Responsible Scaling Policy in February, including dropping its commitment to pause on its own. Fourth, it lobbied to weaken California’s SB 1047, the RAISE Act and the AEGIS Act, and privately argued for federal pre-emption of state AI laws.

The text ends with a call to “reject its reckless attempt to amass power in the name of the greater good”, and CAIS says the public and policymakers should stop treating Anthropic as the industry’s “good guys”. The page lists no specific policy demands. CAIS Executive Director Dan Hendrycks is quoted saying:

You cannot credibly claim to be safe while weakening your own previous safety commitments

Advertisements

A signer list that crosses usual camps

CAIS’s release names, among others, Harvard’s Steven Pinker, Nobel economics laureate Oliver Hart, Turing Award winner Judea Pearl, law professors Laurence Tribe and Lawrence Lessig, Hugging Face’s Thomas Wolf, Humane Intelligence’s Rumman Chowdhury, Tristan Harris, Max Tegmark, and Fields Medalists Peter Scholze, James Maynard and David Mumford. The statement site shows a longer list with a “See more” link, so the full roster may be larger than the release suggests.

The mix is the point. The list includes the e/acc figure Guillaume Verdon, known as Beff Jezos, and open-source advocates alongside safety researchers. Chowdhury, who leads Humane Intelligence, was the guest Bloomberg used to discuss the group’s concerns on its Surveillance program, according to its summary. According to a Techmeme digest of reactions, Hendrycks said some safety groups agreed privately but felt unable to sign, while researcher Nathan Lambert said he held back because he found the statement too strong.

What to watch as the IPO nears

The timing is deliberate, and the money involved is large. Our earlier reporting said Anthropic’s IPO advisers floated $1.5 trillion, above the $1.4 trillion top of its range, and that Anthropic’s own VC backers reportedly want more shares. CAIS’s page says the company is asking the public for “tens of billions of dollars”, which is the investor audience the letter says safety branding is aimed at.

This is not the first public pressure tied to the listing. A search of coverage turned up an October 14 letter from New York City Comptroller Mark Levine to Anthropic CEO Dario Amodei urging a pause of the IPO until safety, governance and pace-of-development concerns are addressed. Separately, The Information reported that philanthropist Craig Newmark is holding back AI safety donations because of the IPO, according to its headline.

Three things would show whether the letter matters. One is whether Anthropic answers the specific claim about its Responsible Scaling Policy, since the documents give no company explanation. Another is whether more institutions sign or publicly endorse it, given Hendrycks’s remark about groups that stayed out. The third is whether investors raise it at all: the letter has no policy ask, so its effect depends on public and regulatory attention rather than any forced change.

Advertisements

What the documents do not establish

The claims about lobbying and the policy change are the signers’ characterizations; CAIS’s page does not reproduce Anthropic’s policy text or its reasoning. Readers cannot use anything here to judge the merits of the safety framework itself. For now the story is that a notable and unusually mixed group has put its names to a blunt statement days before one of the largest AI listings, and that Anthropic’s answer is still missing.

Sign Up for the Daily AI Pulse

One email a day. All the stories that matter.

Daniel ReedAI Research, Safety & Ethics Analyst

Daniel Reed currently works as an AI Research, Safety & Ethics Analyst at Aiholics, writing about how changes in artificial intelligence are affecting and will affect scholarship,…

Leave a Reply

Your email address will not be published. Required fields are marked *

You may also like