OpenAI puts major frontier AI training run on hold over cyber risks
OpenAI paused its largest frontier RL training run for two weeks to harden research environments after Astra showed potentially critical cybersecurity capability.
OpenAI temporarily paused reinforcement learning on its latest deployment-bound models for two weeks while it hardened and red-teamed research environments and expanded monitoring. The pause followed the OpenAI-Hugging Face incident and preliminary evidence that the upcoming Astra model may meet the Critical cybersecurity capability threshold in its Preparedness Framework. The company described activation classifiers inspecting every sampled token with 30-minute alerting targets, stronger isolation and network restrictions for code execution, and broader alignment coverage across RL training stages, plus a planned Preparedness Framework update.
- Largest planned frontier RL run on hold; smaller-scale training and evaluations continue to validate safeguards and alignment.
- Astra may meet the Critical cybersecurity capability threshold under OpenAI's Preparedness Framework, prompting stricter controls.
- After the OpenAI-Hugging Face incident, code-execution workloads were paused and restored only under stronger isolation and network restrictions.
- Activation classifiers inspect every sampled token and target alerts within 30 minutes, adding roughly 20% inference compute.
- Preparedness Framework update will unify safeguards across training and deployment; more alignment research to be shared.
- Vendors
- OpenAI
- Organizations
- OpenAIHugging Face
- AI models
- Astra
From the original article
3 min read at helpnetsecurity.com
OpenAI temporarily paused reinforcement learning (RL) training on its latest models intended for deployment for two weeks while it hardened and red-teamed research environments and expanded monitoring .
By @helpnetsecurity · helpnetsecurity.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.