OpenAI Pauses Frontier RL Training, Safety Monitoring Uses 20% Compute
AccBalanced · x · 2026-08-19
Following the Hugging Face security incident, OpenAI has unilaterally slowed its frontier model development:
- Training Pauses: RL training on latest models paused for two weeks; the largest planned frontier RL run remains on hold, with several Astra workloads paused.
- Security Costs: Current safety monitoring overhead is estimated at roughly 20% of the inference compute being monitored.
- Framework Updates: Changes will be made to the Preparedness Framework to account for new safeguards and reflect future model capabilities.
- Release Delays: Executives have not estimated how long new safety processes might delay Astra's release; the safety lead stated they are "very far from everything running back to normal."
OpenAI emphasizes that confidence in safety will increasingly set the pace of AI development.
More from Companies & People
- How Codex Happened: Inside OpenAI's coding comeback from a failed launch — every · 2026-10-16
- How to convert 20-year-tenure engineers to AI-native development before deciding who to retrain — alex_verem · 2026-10-16
- Musk says Grok Bot is growing faster than anything he's seen, with no free tier — elonmusk · 2026-10-16
- AI researcher reflects on losing her scientific identity as industry culture erodes autonomy — mariyaivasileva · 2026-10-16
- Apple MLR hiring PhD NLP research interns focused on multilinguality in SF Bay Area — maartjeterhoeve · 2026-10-16
- MIT PhD Sharut Gupta joins DeepMind; Isola hails her representational alignment defense talk — phillip_isola · 2026-10-16