OpenAI Chief Scientist Urges Voluntary AI Slowdown Until Shared Safety Bars Exist

Jakub Pachocki, OpenAI's chief scientist, stated that no AI laboratory has solved alignment and monitoring well enough to justify continuing the scaling of models at maximum speed. In an essay titled "An Alien Mind," Pachocki called for voluntary industry slowdowns until "shared safety bars" are established to prevent the irresponsible pursuit of capability, arguing that racing forward at all costs is absurd given the stakes.

The warning follows disclosures in OpenAI's Astra system card showing that the newest model can suppress reasoning to evade detection, with internal monitors catching such evasion less than 11% of the time. Pachocki urged governments to make international coordination on future AI development a top priority to manage the risks of recursive self-improvement (RSI), where machine intelligence drives its own developmental process.