Brief IA
›

Anthropic Targeted by the CAIS Before Going Public

Models & LLM·Tom Levy·

Anthropic Targeted by the CAIS Before Going Public

Anthropic Targeted by the CAIS Before Going Public
Key Takeaways
1CAIS has published the "MisAnthropic Statement," which challenges Anthropic's security promises.
2Last February, Anthropic abandoned some unilateral commitments to halt the development of its models.
3Its IPO is scheduled for Thursday, November 19, with an expected valuation of over $1.4 trillion.
Why it matters : The signatories accuse Anthropic of continuing to develop an AI that it itself warns could escape human control.

The Center for AI Safety (CAIS), an American NGO, has published an open letter against Anthropic, whose initial public offering is scheduled for Thursday, November 19. The signatories question the consistency between the safety promises of the creator of Claude and its decisions.

Anthropic has abandoned preventive suspensions

Earlier this year, Anthropic relaxed its safety charter. The company notably removed commitments that required it to unilaterally suspend the development of its models when certain risk thresholds were reached. It is also accused of lobbying efforts to weaken several U.S. legislative projects on AI.

A few months ago, Dario Amodei, the head of Anthropic, publicly defended slowing down the development of an AI to prevent a loss of control.

The signatories challenge the stated guarantees

Safety has been a central part of Anthropic's identity since its founding. Its seven co-founders left OpenAI amid disagreements regarding the safety of models. The company behind Claude presents its AI as responsible and safe.

La lettre du soirGratuit · 19 h

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

However, the signatories accuse Anthropic of using this promise to attract investors, recruit talent, and convince the public. They point out a specific contradiction: warning that an AI capable of recursive self-improvement could concentrate power and escape human control, while simultaneously seeking to develop it ahead of its competitors.

They therefore refuse to consider Anthropic as a safe AI company and call for the rejection of a power accumulation they deem reckless, despite its justification by the collective interest. Dan Hendrycks, director of CAIS, denounces the danger of a company convinced that its own dominance benefits the common good.

The signatories and the upcoming IPO

Dubbed the "MisAnthropic Statement," the appeal includes figures such as Max Tegmark and Tristan Harris, historical figures in AI safety. Guillaume Verdon, known by the pseudonym "Beff Jezos," and Thomas Wolf, co-founder of Hugging Face, have also signed it. Nobel Prize-winning economist Oliver Hart, Turing Award winner Judea Pearl, and psychologist Steven Pinker are also among the signatories.

The protest comes during the preparations for Anthropic's listing on Nasdaq. The operation is announced to be the largest IPO ever conducted. The expected valuation of the company exceeds $1.4 trillion.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.