Mark Zuckerberg rejects calls to slow AI development, backing independent evaluators instead

Meta chief executive Mark Zuckerberg has rejected calls for an industry-wide slowdown in artificial intelligence development, arguing instead that labs should rely on independent evaluators to ensure their systems are safe.

Writing on social media platform X on Tuesday, Zuckerberg described third-party assessments as "industry best practice". He positioned them as a more practical alternative to wait-and-see pacts between rivals, injecting a sharp counterpoint into a growing global debate over how to prevent advanced AI from harming humanity.

The debate over the economic and safety risks of AI accelerated over the weekend with the publication of a 3,800-word essay by Anthropic chief executive Dario Amodei. In his piece, Amodei urged a slowdown in the pace of development for the most advanced, powerful models, arguing that researchers need more time to identify and comprehend the potential threats they pose.

Amodei's call received high-profile support from rivals who are normally locked in fierce competition. OpenAI chief executive Sam Altman and Elon Musk, the head of SpaceXAI, both endorsed the essay. Furthermore, an OpenAI executive said the company is currently working with Anthropic and Alphabet's Google on collaborative AI safety initiatives.

Zuckerberg, however, argued that safety is best handled by individual companies taking responsibility for their own work, rather than demanding that their competitors halt progress. To illustrate his point, he revealed that Meta had unilaterally decided to delay the release of its "Muse" AI tool for several months to focus on safety and security.

"We didn't call for everyone else to do this before we would," Zuckerberg wrote. "We just did it as part of our day-to-day work because it was clearly the right thing for people and for us."

This disagreement highlights a fundamental split in Silicon Valley over the future of AI governance. On one side, companies like Anthropic, OpenAI, and Google are increasingly discussing collective agreements, joint research, and shared standards to manage risk. On the other side, Meta has championed a more independent and open approach, releasing its models publicly and arguing that safety is best achieved through transparent, external peer review rather than central restrictions.

By advocating for independent evaluators rather than a universal pause, Zuckerberg is promoting a model of safety that allows tech giants to keep building at pace, provided they let external experts review their systems. But with governments around the world drafting hard rules to regulate the technology, it remains to be seen whether voluntary third-party evaluations will be enough to satisfy regulators who are increasingly wary of letting the industry police itself.