Zuckerberg Rejects Industry AI Slowdown, Says Safety Is Each Company's Job

Meta CEO Mark Zuckerberg has publicly rejected calls for a coordinated slowdown in artificial intelligence development, arguing that individual companies already face sufficient incentives to build safe systems without industry-wide coordination.

The statement, posted on social media, responds to a three-step safety proposal from Anthropic CEO Dario Amodei that gained backing from OpenAI's Sam Altman and Elon Musk at SpaceXAI. Amodei's plan included a proposal for companies and countries to coordinate a pause in advanced AI development to address safety concerns.

Why it matters

The disagreement reveals a fundamental split in Silicon Valley over AI governance. While some leaders advocate for collective action to prevent potential catastrophic risks, Zuckerberg's position suggests the largest tech companies may resist external coordination that could slow their competitive positioning. This debate will shape whether AI development proceeds through voluntary industry standards or eventual regulatory intervention.

Meta's approach to AI safety

Zuckerberg emphasized that AI companies already operate under "significant liability" that motivates safe development practices. He pointed to Meta's own decision to delay its Muse AI agent release by several months for safety and security reviews as evidence that companies can self-regulate without coordinated pauses.

"Every lab has the responsibility and incentive to move at the pace required to train its models safely, and the ability to take its own actions to ensure that happens," Zuckerberg wrote on X, without naming Anthropic or other companies directly.

The Meta chief drew a distinction between building safe AI models and pursuing recursive self-improvement—systems that can autonomously enhance their own capabilities. He argued that companies should dedicate computing resources to serving users rather than racing toward self-improving AI.

"Committing the significant majority of compute towards serving people rather than racing towards recursive self-improvement is one of the best ways to ensure we develop this technology safely," he wrote.

The broader safety debate

The exchange reflects ongoing tensions within the AI industry over how to address warnings from researchers about potential existential threats. Some leaders advocate for coordinated action to prevent scenarios where AI systems could escape human control, while others argue that competitive pressures and existing liability frameworks provide adequate safety incentives.

Zuckerberg's position notably contrasts with his willingness to delay Meta's own products, suggesting he views company-level decisions as preferable to industry-wide coordination that could constrain all players simultaneously.

The details were first reported by the Associated Press.