Anthropic and OpenAI want AI companies to slow down, Mark Zuckerberg has other plans

Mark Zuckerberg has chimed in on the debate around AI safety. Unlike Sam Altman or Dario Amodei, the Meta CEO does not want regulation to slow down AI development. Instead, he believes every company must be responsible on their own.

In Short

• In an X post, Zuckerberg called independent evaluators an industry best practice
• He said Meta delayed Muse for months over safety and security
• Zuckerberg argued harmful models face liability, creating incentives for caution

OpenAI and Anthropic want to slow down AI development. Recently, Anthropic’s Dario Amodei has suggested ways through which companies can work together to achieve this, something that his rival Sam Altman has also agreed upon. But it seems that Meta CEO Mark Zuckerberg is in no mood to follow such an approach. Instead, Zuckerberg believes every company should already ensure that their models remain safe.

In a post on X, Mark Zuckerberg joined the debate surrounding the future of AI. The Meta CEO insisted AI companies do not need a framework to build safe AI, but do it by themselves. “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,” he said.

Zuckerberg went on to claim that Meta had already delayed the release of its Muse AI agent based on safety concerns without having to ask others to slow down. “We didn't call for everyone else to do this before we would,” he wrote. “We just did it as part of our day-to-day work because it was clearly the right thing for people and for us.”

Mark Zuckerberg’s remarks place him at odds with Sam Altman and Dario Amodei, who have urged for some sort of regulation to help keep check on AI development – something which even Elon Musk seems to agree with. The fears around slowing down AI have become the centre of debate after former Anthropic AI researcher Jacob Coxon claimed that AI could potentially end humanity.

Mark Zuckerberg wants no US policy on AI slow down

In the same post, Zuckerberg linked his essay from last month where he discussed that any US policy would hurt US AI labs. “Any policy that slows American model releases -- even by a month -- could add significant risk to American leadership while letting foreign models race ahead,” he wrote at the time.

While this puts him against leaders like Altman and Amodei, it also aligns him with US President Donald Trump. Trump has rejected the idea of slowing down American AI research amid fears of losing the AI war to China. “Whoever wins AI, wins,” he said recently. In a conversation with Donald Trump, Nvidia CEO Jensen Huang agreed with such a view. “You’re right. We’re not going to let that happen, sir,’” he said.

While former White House AI czar David Sacks has also hit out at OpenAI and Anthropic, stating that the two companies could slow down AI research on their own. “You guys are the frontier. By any reasonable metric — market share, revenue growth, model capability — the two of you have a duopoly on frontier intelligence,” he wrote on X.

Coming back to Mark Zuckerberg’s X post. The Meta CEO wrote that safety was important for all AI companies today as they could be held liable if things go wrong. “Labs face significant liability if their models cause harm, so they have a strong incentive to prevent this as well,” he added.

But Zuckerberg accepted that “engaging independent evaluators and advisors” was a good move. According to the Meta chief, the company’s AI division – Meta Superintelligence Labs (MSL) – already did this in several areas because it helped produce better work. “I believe the key to building a positive future for everyone is maintaining the right balance of power,” he wrote.

This post was Zuckerberg's first public intervention in the latest round of arguments over how to address rising concerns about AI safety. Apart from comments regarding how to slow down AI research, Anthropic CEO Dario Amodei has also accepted that the industry may have lied about risks related to AI. At the same time, Anthropic’s safety researcher Evan Hubinger has predicted that AI may end humanity within the next decade. OpenAI’s Marcus Williams, on the other hand, says that such a doomsday scenario may happen within the next three years.

- Ends