Meta CEO Mark Zuckerberg says individual AI companies should be responsible for deciding how quickly they develop their technology, distancing Meta from recent calls for an industry-wide slowdown over safety concerns.
His comments came after Anthropic CEO Dario Amodei proposed greater coordination between AI companies and governments to slow advanced AI development while safety measures catch up.
OpenAI CEO Sam Altman and Elon Musk have also expressed support for greater coordination around AI safety.
Zuckerberg argued that AI companies already have strong incentives to develop their models safely because they face significant liability if their systems cause harm. He said each AI lab should determine the pace required to safely train and release its own models.
As an example, Zuckerberg said Meta delayed the release of its Muse AI agent for several months while working on safety and security issues. Meta made that decision independently rather than asking competing AI companies to pause their own development.
Zuckerberg also argued that companies should prioritize AI systems designed to serve people rather than focusing their resources on recursive self-improvement, where AI systems become increasingly capable of improving themselves.
His position highlights a growing divide among major AI companies over how the industry should respond to increasingly powerful systems. While executives including Amodei and Altman have advocated greater coordination and slower development under certain circumstances, Zuckerberg favors allowing individual companies to manage their own pace and safety measures.