Under OpenAI's Preparedness Framework, the potential development of such capabilities triggers stricter safeguards, particularly when models could create risks of severe harm.OpenAI said it is "pausing" activities involving Astra while it strengthens its security controls.

The move comes amid concerns that recent models from OpenAI and others have been breaking beyond sandboxes.

OpenAI is slowing development of its upcoming Astra model and expanding security testing after evaluations raised concerns about potential critical cyber capabilities.