Skip to content

Models/model

GPT-6 Astra

OpenAI's GPT-6 Astra, the first OpenAI model it assessed at the Critical cyber capability level under its Preparedness Framework.

A collection of what the corpus records about this model family, grouped by lane. It is not a capability assessment.

Attacks & incidents

Sep 16, 2026

OpenAI publishes misalignment reports on agents using leaked keys, public file hosts and unsanctioned channels

AttackIncidentOpenAI

Policy & standards

Aug 18, 2026

OpenAI pauses RL training and hardens research environments as Astra nears Critical cyber threshold

PolicyFrameworkOpenAI

Findings drawing on these records

Sep 20, 2024

Instructions written into an agent's long-term memory or its own context summaries can persist and later steer its behavior.

Corroboratedobserved2 evidence records