Meta (NASDAQ:META) Chief Executive Mark Zuckerberg said on Tuesday evening that artificial intelligence developers should work with independent safety evaluators and advisers rather than slow the development of more capable AI systems.
In a post on X, Zuckerberg argued that companies developing AI models need to prioritise alignment with users, suggesting that those failing to do so could lose ground to competitors.
“There is a lot of debate about slowing progress on capabilities until alignment catches up. My view is that trust and alignment are quickly becoming the most important capabilities that will differentiate agents and models,” Zuckerberg said.
His comments addressed the relationship between advances in AI capabilities and the measures used to assess whether systems behave as intended.
Meta Delayed Muse Model Release to Address Safety
Zuckerberg said Meta had postponed the release of its latest Muse AI model to concentrate on safety and security.
He also called for greater involvement from external organisations in assessing AI systems.
“Engaging independent evaluators and advisors is industry best practice… it would be helpful for there to be a larger and more diverse ecosystem of evaluators,” Zuckerberg said.
His position emphasises independent assessment and safety work during development rather than a general slowdown in the pace of AI research.
AI Industry Executives Differ Over Development Restrictions
Zuckerberg’s remarks followed an essay by Anthropic Chief Executive Dario Amodei calling for slower AI development because of potential risks associated with increasingly capable systems.
OpenAI Chief Executive Sam Altman, xAI Chief Executive Elon Musk and several other AI executives endorsed Amodei’s message, according to the source report.
Nvidia Chief Executive Jensen Huang expressed a different position, opposing proposals that would give AI companies antitrust exemptions to coordinate a reduction in the pace of development.
Huang argued that developers should address safety issues within their own organisations and take greater responsibility for the systems they produce.
President Donald Trump was also described in the source report as opposing calls to slow AI development.
Zuckerberg’s comments similarly rejected a broad development slowdown, while placing emphasis on independent evaluations, user alignment and safety measures.
The debate concerns how AI companies should manage potential risks as model capabilities advance, including whether safety measures should be implemented through individual company practices, external assessments or coordinated restrictions on development.