OpenAI has previewed the safety precautions it is implementing ahead of releasing Astra, its latest large language model with notable cybersecurity capabilities. The model has drawn attention for being exceptionally skilled at exploiting computer systems, raising both interest and concern in the security community.
Background
OpenAI has increasingly positioned its models around both capability and safety testing, with cybersecurity becoming a key evaluation domain for frontier AI systems.
- Source
- TechCrunch
- Published
- Sep 2, 2026 at 05:06 AM
- Score
- 7.0 / 10