On September 15, Meta CEO Mark Zuckerberg rejected calls for AI labs to coordinate a slowdown in frontier model development. Days earlier, Anthropic CEO Dario Amodei had proposed that leading AI companies pace their progress together, with support from OpenAI's Sam Altman and Elon Musk.
Zuckerberg said each lab already faces "responsibility and incentive" to move safely. He pointed to Meta's Muse AI agent, which the company delayed for several months internally without asking competitors to wait. Market forces will handle safety, he argued: users won't pick agents that ignore instructions, so labs that skip safety work "will fall behind."
Who stayed in, who walked out
The slowdown camp now includes Amodei, Altman, Musk, and Google DeepMind's Demis Hassabis. Against: Zuckerberg, Nvidia CEO Jensen Huang, Donald Trump, and Beijing. A coordinated pause is a prisoner's dilemma. It collapses the moment one player defects — and several already have.
The implication: If you rent AI from these companies, they set the speed. You get no vote on whether the next model ships in three months or three years. The race is back on, and the finish line keeps moving.