The company says a new system crossed its internal threshold for high-risk cyber capabilities.
OpenAI has reportedly delayed the release of a new AI model after internal safety evaluations flagged advanced cyber capabilities. The company says the model was not involved in a recent Hugging Face breach.
The pause shows how frontier AI labs are increasingly treating model launches as security events, not just product releases. If systems demonstrate cyber capabilities beyond internal safety thresholds, companies may face pressure to delay deployment, tighten safeguards, and explain their risk decisions more publicly.
Set tight time constraints for a task to increase efficiency
This prompt has 3 customizable fields.
Join The Vault for full access to all prompts, custom GPTs, and more.