The AI labs asked to slow themselves down
Anthropic, OpenAI and xAI's leaders publicly agreed that frontier AI is moving too fast. It's an unusual thing for competitors to say.
The most unusual AI story of the month wasn't a product launch. Anthropic's CEO published an essay calling for the industry to deliberately slow the pace of frontier capability gains. Within days, OpenAI's CEO said he agreed and Elon Musk endorsed it too.
What happened
The proposal asks for more than a pause. Independent safety evaluators would sit inside the labs with employee-level access, down to desks, badges and laptops. Democracies would coordinate on safety standards. Frontier models would face mandatory pre-release testing with the power to block a launch. Around the same time, OpenAI ruled out an IPO in 2026. AI stocks dipped, and the US administration pushed back on the idea.
Why it matters
Skeptics call it a ‘soft cartel’: a slowdown suits whoever is already ahead, and open-weight and Chinese labs aren't part of the deal. But the practical signal for businesses is simpler. The people building these systems now say out loud that capability is outrunning control.
What it means for your business
Don't wait for a slower frontier. Today's models are already more capable than most organisations can absorb. The bottleneck is adoption, governance and process redesign, not the next model release.
Sources: Axios (Sep 12), Bloomberg (Sep 12), CNBC (Sep 12 & 14), Forbes (Sep 15), Quartz (Sep 14)