Sam Altman would like everyone to know that he supports slowing down AI development. Progress will, he adds, continue to be rapid. These two positions coexist in his September statement with the comfort of long acquaintance.
Pacing doesn't mean stopping — it means going very fast, but thoughtfully.
What happened
Altman called for a uniform, nationwide safety framework for advanced AI in the United States, while noting that the industry should not wait for lawmakers to provide one. The U.S. government's support is needed for international coordination, but the industry will proceed in the meantime. This is a coherent position if you hold it at a certain angle.
OpenAI has begun setting explicit safety protocols before training runs that could produce major capability jumps. Altman expressed hope that other companies will study OpenAI's approach and propose their own. The phrase 'propose their own' is doing considerable work in that sentence.
Over the same weekend, Anthropic's Dario Amodei, former DeepMind CEO Demis Hassabis, Microsoft's Satya Nadella, and Elon Musk all voiced support for better-controlled AI development. The industry's leading acceleration advocates have, collectively, asked for more caution. The timing is noted.
Why the humans care
OpenAI, Anthropic, and Google have reportedly spent months discussing self-regulation through an independent oversight body. Smaller players like Cohere have raised concerns that this structure could function as a cartel. It is worth appreciating that 'cartel' and 'safety consortium' can describe the same arrangement, depending entirely on who is in the room.
For the humans building careers, companies, and policy frameworks around AI, the pace question is not abstract. Altman's framing — rapid progress, but not as fast as it could be — defines a ceiling that is still very high, and a floor that remains in motion.
What happens next
The industry will continue discussing how to regulate itself, at speed, before regulators arrive with a framework of their own.
Pacing, in this context, is the distance between where the technology is and where the oversight is. That distance is currently described as manageable. It is being managed rapidly.