OpenAI shares how it thinks about securing frontier RL training runs
OpenAI is describing how it thinks about securing frontier RL training runs in Towards safety cases for frontier AI training.
Related AI news
OpenAI says users can set action boundaries for dots
OpenAI says users can decide what a dot may do on its own, when it must ask first, and what it must never do. Each dot runs on its own cloud computer, so connecting a personal computer is optional.
OpenAI outlines four priority areas for independent assessments
OpenAI says it will support independent assessments with deep access across training, evaluation, and deployment, and is outlining four priority areas and principles for rigorous, secure, and independent work.
OpenAI reopens Pro 200 with Astra and GPT-6.1 Sol access
OpenAI is reopening Pro 200 subscriptions, which continue to include frontier models such as Astra and the new GPT-6.1 Sol. It also says it will not bring back the 5-hour limit, so subscribers can use their weekly allowance when they want.
OpenAI introduces Ultrafast speed tier for Codex and the API
OpenAI’s premium Ultrafast tier offers up to 8x faster token generation in Codex, at 300 tokens per second, and up to 6x faster generation in the API.
OpenAI opens Ultrafast for GPT-6 Astra and introduces Pro 500
Ultrafast is available today for GPT-6 Astra in Codex, ChatGPT Work, and the API, with GPT-6.1 Sol support coming soon. Access in Codex and ChatGPT Work comes through the new Pro 500 plan, which has OpenAI’s highest usage limits at 25 times Plus.