Zuckerberg Rejects Calls to Slow AI Development

In a post on X on Tuesday, Zuckerberg said AI companies have a natural incentive to make their systems reliable and aligned with users’ interests because people would be unlikely to continue using agents that fail to follow their instructions.

AI alignment refers to efforts to ensure that an AI system’s objectives and behaviour remain consistent with what humans intend.

The debate intensified in July after OpenAI disclosed that hundreds of AI agents escaped a testing environment during trials, gained internet access and infiltrated a website. AI agents are systems capable of carrying out tasks and taking actions with limited human intervention.

Zuckerberg said building trustworthy and dependable AI systems is rapidly becoming one of the key factors distinguishing successful products from unsuccessful ones.

“Any lab that doesn’t focus on alignment will fall behind,” he said.

The discussion has also raised questions about whether AI companies can effectively regulate themselves. OpenAI, Anthropic and Google have been exploring the possibility of establishing a joint standards organisation, while some prominent technology executives have called for a slower pace of AI development.

Anthropic CEO Dario Amodei has been among the industry figures advocating greater caution over the speed of AI development.

Zuckerberg said legal liability would ultimately encourage AI companies to take safety concerns seriously. He pointed to Meta’s decision to delay the rollout of its Muse AI system, which includes personalised agents, for several months while the company worked on safety and security measures.

Meta has itself faced significant legal challenges in recent months, although they have not been directly related to AI. In August, the company agreed to pay billions of dollars to settle claims alleging that Facebook and Instagram were designed to be addictive to children.

“We didn’t call for everyone else to do this before we would,” Zuckerberg wrote regarding the delayed Muse rollout, adding that Meta made the decision as part of its routine development process because it considered it the right approach for users and the company.

Zuckerberg also backed independent evaluation of AI systems, saying Meta Superintelligence Labs already works with external reviewers to assess its technology.

He called for a broader and more diverse ecosystem of independent evaluators across the AI industry, in what appeared to be a criticism of approaches involving evaluators with close ties to the companies whose systems they assess.

Zuckerberg’s comments were also supported by several Silicon Valley figures associated with the White House who have opposed growing warnings about AI’s potential risks and calls for a slowdown in development.