Mark Zuckerberg says AI does not need an industry-wide slowdown because competition will force companies to build safer models.
In a social media post Tuesday, the Meta CEO said people will not want to use agents that are misaligned with them or that do not do what they ask. That, he wrote, gives frontier AI labs a 'strong natural incentive' to make their models more aligned.
Zuckerberg said Meta delayed the release of Muse, the AI agent it launched earlier this month, because of safety and security concerns. He said that decision showed some companies are already self-policing on AI safety.
'We didn’t call for everyone else to do this before we would. We just did it as part of our day-to-day work because it was clearly the right thing for people and for us,' he said.
His comments place him on the side of Nvidia CEO Jensen Huang, who has rejected calls from Anthropic CEO Dario Amodei and others for an industry-wide slowdown. At Salesforce’s Dreamforce event Tuesday, Amodei compared AI safety concerns to those faced by the automotive industry and argued companies should examine their own operations after a safety incident.
Amodei also said that even with a theoretical freeze on AI, companies are still 'making use of maybe only 5% or 10% of what the possible value of the technology is.'
Huang took the opposite view, saying safety should remain a top priority but that companies can solve the problem themselves. He said if a company notices a product is not safe, it should pause and fix it, but otherwise companies should 'run as fast as they can.'
'Innovation, speed, and safe products—it’s a false choice,' Huang said. 'You could definitely have both at the same time.'
Zuckerberg said Meta is committing a significant majority of its computing power toward serving people rather than racing to develop AI systems capable of recursively improving themselves.
He also said, 'I believe the key to building a positive future for everyone is maintaining the right balance of power. This is within our power to do.'
The bigger issue is whether the market can discipline AI faster than regulators can. Zuckerberg is betting that customers, competition, and corporate incentives will do what bureaucrats want to do through rules. That is a familiar case for free enterprise: let capital reward the firms that build products people actually trust.
But the stakes are not small. Recent security incidents and growing calls for regulation show why the debate is now moving from theory to risk management. The market has already voted on speed; what remains is whether it can also enforce safety before the next mistake becomes expensive.
