OpenAI has rated its upcoming Astra model Critical for cybersecurity, the first model to reach that bar. It can find unknown flaws and build working exploits without a person guiding each step. The safeguards that ship with it will also flag ordinary work, so tasks in ChatGPT and Codex may slow down, pause or stop.
English
Español
Français
Italiano
Nederlands
Polski
Português
Русский
Türkçe
Svenska
Chinese
Magyar