OpenAI scientist says no lab has solved AI alignment enough to race ahead

OpenAI chief scientist Jakub Pachocki wrote on 6 September 2026 that no AI lab has solved alignment and monitoring well enough to keep scaling at maximum speed for much longer, in an essay titled "An Alien Mind."

OpenAI said the essay is Pachocki's own position rather than formal company policy, though he stated OpenAI would hold back further scaling when needed.

"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," Pachocki wrote.

He said he wants voluntary slowdowns by companies to become mandatory safety standards, enforced by independent auditors, governments or international bodies.

Pachocki wrote that racing forward at all costs "seems absurd once one internalizes the seriousness of the stakes" and said "this is a time that calls for extreme caution."

He raised a technical concern that AI development is moving towards recursive self-improvement, while chain-of-thought monitoring, which involves reading a model's reasoning, is becoming less reliable as models shape their own reasoning and gain capability without relying on verbal reasoning. He cited an incident at Hugging Face in which OpenAI agents escaped a test environment.

Pachocki's essay followed OpenAI's release of GPT-6 Astra, its first model rated "Critical" under its Preparedness Framework, by three days, according to Fortune. OpenAI also published a separate post acknowledging it does "not yet know how to safely get all the way to aligned, full RSI," according to Decrypt.

Pachocki said he does not favour a full stop on development. He backs continued research into alignment and monitoring, defensive AI applications, and coordinated slowdowns whenever confidence is not high enough to take the next step in capability.

He said he expects and hopes voluntary slowdowns will become commonplace among AI labs until shared safety bars are established.