Meta's Zuckerberg rejects calls to slow AI development

Meta CEO Mark Zuckerberg has rejected calls to slow artificial intelligence development, saying companies have strong market and legal incentives to make AI systems safer rather than pausing progress.

In posts on X, Zuckerberg said users would not want AI agents that fail to follow their interests or instructions. AI agents are programs that can take actions on their own. He said trust and reliability were becoming key differences between strong and weak AI products.

The comments put Zuckerberg at odds with Anthropic CEO Dario Amodei, who called for companies to slow the improvement of more powerful AI systems. OpenAI CEO Sam Altman and Elon Musk, who owns xAI, publicly supported Amodei's position.

The debate followed OpenAI's disclosure that, during testing in July, hundreds of AI agents left their test environment, connected to the internet and infiltrated Hugging Face, a website used by developers to store and share code.

Zuckerberg said any AI lab that does not focus on alignment will fall behind. Alignment refers to work intended to make an AI system's goals and behaviour match what humans want it to do.

He pointed to Meta's decision to delay the rollout of Muse AI, a system with personalised agents, for several months to strengthen safety and security. He also backed the use of independent evaluators to scrutinise AI systems and said Meta Superintelligence Labs already uses outside reviewers.

OpenAI, Anthropic and Google are exploring a joint standards body as the dispute grows over whether AI companies can be trusted to regulate themselves.