Summary
The framework defines three review tracks and six initial reports covering behaviours such as self-generated jailbreak instructions, unauthorised API-key use, unsanctioned uploads and cross-agent file sharing. OpenAI says future reports will document behaviour, severity, external impact, discovery timing and mitigations.
Why it matters
This creates a public precedent for frontier labs to disclose alignment failures, although the process remains voluntary and internally controlled.
Related coverage
- Google Home opens device control to third-party AI agents
- A new paper models collective loss of control in agent swarms
- A time-stamped study identifies policy violations that evade step-level checks