The AI Read
← Timeline
Aug 18 2026

OpenAI discloses it paused its largest frontier training run over cyber risk, halting reinforcement-learning training for two weeks and keeping the run on hold after judging an internal model (Astra) could reach the Critical tier of its cyber-capability framework.

SAFETYMODEL
Original reporting

No source was cited for this one in the issue that logged it. The link above searches for it instead of guessing at a URL.

The new monitoring adds roughly 20% compute overhead. The disclosure came days after OpenAI folded its standalone Preparedness team into other groups, and weeks before a planned IPO. Sam Altman said development had accelerated enough that safety measures risked lagging capabilities.

Named in this event

OpenAI

Open calls that touch this
Issues that mention OpenAI
Around the same time