AI · Security · 2026-11-08 09:42
Frontier Labs evaluates the real-world capabilities of models like GPT-6 Astra in reproducing and exploiting vulnerabilities, using benchmarks such as Cybench and CyberGym to assess their effectiveness in AI-driven cybersecurity tasks.
Auto-generated summary for reference only. RSS