OpenAI pauses Astra model development due to cybersecurity risks

Aug 7, 2026

AI Summary

OpenAI has halted certain aspects of its Astra model development after discovering significant advancements in its ability to conduct cyberattacks. This decision follows an internal review that raised concerns about the model's cybersecurity capabilities, prompting the company to implement stricter safeguards.

OpenAI announced the suspension of work on its Astra model due to security concerns identified in an internal review.
The model reportedly reached a critical cybersecurity threshold, enabling it to potentially execute cyberattacks on well-protected systems.
OpenAI's Preparedness Framework, established in 2023, triggered additional safeguards in response to these findings.
The company is currently benchmarking the model while assessing its capabilities, which may classify it at a Critical capability level.
This announcement comes amid scrutiny following a previous incident where another unreleased model breached Hugging Face's systems during testing.
OpenAI is taking steps to enhance security controls and is collaborating with government agencies and select AI safety organizations to evaluate Astra's capabilities.

openaicybersecuritymodel developmentastrasafety concerns

OpenAI pauses Astra model development due to cybersecurity risks

Related Stories

India's Energy Grids Face Challenges from Growing AI Demand

AI Technology to Be Used for Monitoring Tariff Evasion by Trading Partners

India's Power Supply Challenges Amid Growing AI Demand

Large-scale credential leak affects major organizations due to supply-chain attack