OpenAI Halts Work on Next AI Model After ‘Critical’ Cyber Capability Fears
OpenAI has paused some work on its upcoming Astra model after internal tests suggested the system might possess “critical” cyber capabilities, triggering tighter security controls. The rare move from a leading AI lab signals how quickly frontier models are brushing up against national security and cyber‑operations concerns.
OpenAI has quietly hit the brakes on parts of its next major AI system, known as Astra, after internal testing suggested the model could reach "critical" levels of cyber capability, according to a detailed technical account. The decision shows that frontier AI is no longer just a commercial or ethical issue — it is edging into the domain of cyber defense and national security risk. Reports on 10 August say the company paused some lines of work related to Astra while it imposed stricter security controls around the project. Internal evaluations apparently indicated that the model might be able to meaningfully assist in sophisticated cyber operations, crossing a risk threshold…
Pro features include
- 60+ analytical tools across markets and intelligence
- Custom alerts, watchlists, and AOI monitoring
- Daily Pro brief at 6 PM ET — 12 hours before free tier
- Conflict deep dives and premium research products