The chief scientist of OpenAI has written that no artificial intelligence lab — including his own — has done enough work on safety to justify building bigger systems as fast as it can for much longer, and says he hopes companies start slowing down voluntarily.
“Currently I believe that no lab has solved alignment and monitoring to a sufficient degree to continue responsibly scaling at maximum speed for much longer.”
Jakub Pachocki set that out in an essay titled “An Alien Mind”, published by OpenAI on Sunday. “I expect and hope for voluntary slowdowns to become commonplace until shared safety bars are established,” he wrote.
The most concrete admission in the piece concerns how OpenAI checks what its models are doing. The company has leaned on reading a model’s chain of thought — the working it shows on the way to an answer — as a way of catching bad behaviour. Pachocki writes that this is getting harder:
“Unfortunately our evaluations indicate our ability to rely on CoT monitoring is progressively diminishing.”
He argues the industry’s existing voluntary commitments are not enough on their own. Frameworks like OpenAI’s Preparedness Framework and Anthropic’s Responsible Scaling Policy should, he writes, become “widely mandated safety bars for continued development”, and “can be enforced by a network of third-party auditors, by government agencies or by international bodies.” He adds that “international coordination on future AI development needs to become a top priority for governments around the world.”
The essay says nothing about Canada, and names no country’s rules. It is one company’s chief scientist writing on that company’s website, not a regulatory filing or a commitment to any deadline — and it does not say OpenAI is slowing down.
Source: “An Alien Mind”, Jakub Pachocki, OpenAI, 6 September 2026.
