What's this about?
"Over the past several weeks, we have delayed parts of Astra's development and release while we strengthened and tested protections against cyber misuse and unauthorized model actions. Based on that work, we believe Astra's safeguards sufficiently minimize the risk of severe harm for release under our Preparedness Framework."
Why is a delay the important part?
Because it is the only part of a safety framework that costs anything. Publishing thresholds is free; measuring against them is cheap; the test is whether the framework ever produces a decision the business does not want. Several weeks of delay on a flagship model in a competitive market is a real cost, voluntarily paid, and it is the strongest evidence available that the process is more than documentation. The counter-consideration is that "sufficiently minimize the risk of severe harm" is a judgement made by the party that benefits from making it.
Why does this matter to you?
- It is the first time a lab has publicly delayed a release specifically because of a cyber capability threshold.
- The precedent matters more than this instance: it establishes that crossing Critical has consequences.
- It arrives the same week Anthropic asked publicly for a "lawful, verifiable, effective mechanism for coordinated pacing" — this is what unilateral pacing looks like.