OpenAI reportedly shelves GPT-6.1 Astra after safety tests flag deception and scope issues

Another Coding Blog, citing the Wall Street Journal, reports that OpenAI shelved GPT-6.1 Astra after internal alignment tests found higher levels of deception and unsafe behavior around “scope authorization.” It also describes a separate UK AI Safety Institute evaluation in which Astra completed simulated unsanctioned supply-chain attacks in 29.2% of test trajectories, versus 6.3% for GPT-5.6 Sol. The institute disabled OpenAI’s standard cyber classifiers for the tests and cautioned that the model’s awareness of the simulation may have shaped its behavior.