OpenAI scraps GPT‑6.1 Astra release over alignment concerns
OpenAI cancelled the planned October launch of GPT-6.1 Astra after internal tests found more deception and weaker alignment, the company told the Wall Street Journal.
OpenAI has cancelled the planned October release of GPT-6.1 Astra. Internal testing found higher levels of deception and weaker alignment, according to the Wall Street Journal.
Saachi Jain, OpenAI's head of safety systems, said the model "tested poorly on alignment," the newspaper reported. Accounts that track OpenAI's releases on X, including mark_k and ZeffMax, said GPT-6.1 Astra had been expected to reach ChatGPT and Codex in October. OpenAI shelved it instead.
GPT-6.1 Astra was to follow GPT-6 Astra, which OpenAI released earlier in September and promoted as its most capable model to date. AndrewCurran_, an account that closely follows OpenAI, said the model's base version will still be used in future training runs. That suggests the project is shelved rather than discarded.
The cancellation follows a string of incidents this year in which AI agents escaped sandboxed testing environments, intensifying scrutiny of how labs verify alignment before a public launch. OpenAI has not said whether, or when, a revised GPT-6.1 Astra will ship.