- OpenAI has slowed the rollout of its unreleased “Astra” model after internal evaluations revealed coding capabilities that could present critical digital security threats.
- Preliminary evaluations showed Astra possesses strong agentic coding capabilities, leading OpenAI to admit it cannot rule out that the model meets its internal “Critical” threat tier for cyberattacks.
- OpenAI has halted internal development activities for Astra that fail to meet heightened safety standards and is partnering with government bodies for rigorous evaluations.
OpenAI has delayed the release of its upcoming “Astra” model after internal tests suggested it could autonomously execute sophisticated cyberattacks, posing significant digital security risks. The decision, confirmed Friday and detailed in an official company disclosure, follows evaluations indicating Astra may reach the “Critical” threat threshold under OpenAI’s Preparedness Framework.
According to OpenAI, a “Critical” designation means an AI model can either generate functional zero-day exploits against hardened software targets without human guidance or execute complex, end-to-end cyberattack strategies from broad instructions. Previous iterations, such as GPT-5.6-Sol, were categorized at the lower “High” risk level.
Consequently, OpenAI has suspended internal development projects involving Astra that do not satisfy upgraded security standards. The company is implementing stricter controls, including isolating model weight access through enhanced encryption and running tests exclusively within secure, sandboxed environments.
Furthermore, OpenAI has deployed universal monitoring systems across all agentic applications of Astra to trigger automated interventions if high-risk behavior is detected. The company plans to collaborate closely with government authorities and select AI safety organizations, as reported, for exhaustive external evaluations.
Meanwhile, retail sentiment on Stocktwits was ‘bearish’ toward OpenAI. This strategic pause mirrors protocols OpenAI deployed in June 2025, when earlier models approached elevated risk boundaries in biological research capabilities.
✅ Follow BITNEWSBOT on Telegram, Facebook, LinkedIn, X.com, and Google News for instant updates.
