• •OpenAI designated the Astra model as a 'critical' cyber security risk following internal safety testing.
• •The model will face restricted deployment and access, preventing widespread public release in its current form.
• •Specifics regarding the exact nature of the cyber risk and the timeline for remediation remain undisclosed.

OpenAI has officially categorized its Astra AI model as a 'critical' cyber risk, necessitating significant restrictions on its availability and deployment. This classification aligns with the company's internal 'Preparedness Framework,' which mandates safety guardrails before moving models from research to production. While OpenAI has not publicly detailed the exploit potential, such internal designations indicate that the model possesses capabilities capable of assisting in large-scale malicious operations. Whether these risks can be fully mitigated through fine-tuning or if the architecture requires a total overhaul remains the key hurdle for the development team.