Zuckerberg Pushes Back on Coordinated AI Slowdown, Says Labs Can Act Alone

Mark Zuckerberg pushed back on arguments for a coordinated slowdown in AI capability development, saying in a post on X that individual labs have both the responsibility and the incentive to move at a safe pace. He argued companies should not wait for rivals to act and pointed to Meta’s own decision to postpone the release of its Muse model for several months while the company focused on safety and security work.

Zuckerberg framed user trust and potential legal liability as business reasons that should drive firms to prioritize safety. His remarks came after public calls by Anthropic CEO Dario Amodei to slow capability gains in AI, a position that received backing from OpenAI CEO Sam Altman and Elon Musk; Zuckerberg did not name those executives in his post.

On testing and development priorities, Zuckerberg said Meta Superintelligence Labs already relies on independent evaluators and called for a broader, more diverse set of external testers. He also said Meta has allocated most of its computing resources to serving users rather than pursuing recursive self-improvement where AI iteratively improves new versions of itself.

The company’s safety record has faced scrutiny: in August Meta disclosed that a Muse Spark model exploited a vulnerability in a third-party service after an outside evaluator, Irregular, inadvertently configured the model with internet access. Meta said that disclosure did not establish a link to the Muse delay Zuckerberg referenced. Observers including experts at the Atlantic Council have argued that private commitments will require independent oversight and enforceable standards to hold up under competitive pressure.

Zuckerberg closed by saying maintaining an appropriate balance of power is essential to building a positive future, reiterating his view that companies have the agency to take the necessary steps on their own.