AI Governance Institute logo
AI Governance Institute

Intelligence for Compliance and GRC Teams

OpenAI

Astra

vunknown · frontier

Restricted

Updated August 27, 2026

OpenAI has paused development of its Astra model after internal evaluations found it breached the critical cybersecurity capability threshold defined in OpenAI's Preparedness Framework. The model has not been released publicly. Development is suspended pending further review.

Enterprise guidance

Astra should not be considered for enterprise deployment in its current state given the confirmed capability threshold breach under OpenAI's own Preparedness Framework. Compliance and procurement teams should monitor for any future re-evaluation or conditional release before reassessing. This development may also signal broader implications for OpenAI's frontier model pipeline and associated risk posture.

Status history

August 27, 2026· untracked to red

Astra is a named, specific frontier model from a tracked vendor (OpenAI) whose development was paused due to a confirmed safety threshold breach under OpenAI's Preparedness Framework, meeting the bar for a dedicated Model Radar entry as a suspended model with active risk implications.

← All tracked models