tech
Exclusive: OpenAI slows release of Astra model citing cyber capabilities
The move comes amid concerns that recent models from OpenAI and others have been breaking beyond sandboxes.
TL;DR
- OpenAI cannot rule out "critical" cyber capabilities in its upcoming model, Astra.
- The company is expanding safety testing and pausing non-essential internal activities for Astra.
- Astra's development will be slowed until stricter security requirements and safeguards are met.
- This is a notable instance of a frontier AI lab potentially delaying a model's progress due to cyber concerns.
- OpenAI is implementing stricter security controls, including isolated testing environments and universal monitoring for Astra.