What's this about?
The public Astra refuses proof-of-concept exploit creation. But OpenAI says Daybreak — its vetted-access security program — will get "less restrictive safeguards in the coming weeks" for PoC validation, malware analysis, and detection engineering.
What exactly is Daybreak?
OpenAI's trusted-defender tier: vetted security professionals get model capabilities the public doesn't. All three labs shipped the same architecture within days — Anthropic's Mythos, Google's Fairwind Program, OpenAI's Daybreak. Train the capability once, gate access by vetting.
Why is this worth noting?
The industry has quietly standardized "dual-use capability behind identity verification" as the answer to offensive-security-capable models. The open question the gating can't answer: vetting decides who gets the tool, not what they do with it afterward.