The latest fight over artificial intelligence began with alarm inside the industry. After an Anthropic researcher resigned last week, warning that AI companies were “gambling with our lives,” Anthropic chief executive Dario Amodei called on Saturday for a global slowdown in development. OpenAI’s Sam Altman, Google DeepMind’s Demis Hassabis and Elon Musk subsequently joined the broader push for coordinated restraint.¹

That emerging coalition sees the pace of frontier-model development as a problem no single lab can solve alone: if one company slows while another races ahead, the safety-minded firm risks being punished for caution. But critics, including technology investor and Trump adviser David Sacks, have questioned whether warnings from Anthropic and other leading labs also serve to entrench their competitive position.¹

On Tuesday, Zuckerberg made Meta’s break with the collective-pause camp explicit. “Every lab has the responsibility and incentive to move at the pace required to train its models safely,” he said, insisting Meta had already acted on that principle.² His message was not that safety concerns are imaginary, but that an industrywide mandate is the wrong tool.

Meta points to its own conduct as evidence. Zuckerberg said the company delayed the release of Muse, its AI agent, for several months to address safety and security — and did so without demanding that competitors halt their work first. “We just did it as part of our day-to-day work because it was clearly the right thing for people and for us.”David Sacks (DavidSacks)³

The divide is also geopolitical. Zuckerberg has warned that even a one-month delay in American model releases could threaten U.S. leadership while foreign models continue advancing.² With Meta preparing its more powerful Watermelon model, the argument is no longer simply whether AI should slow down, but who gets to decide when the brakes are applied.