AIAxios Techabout 2 hours ago
AI outpaces existing benchmark tests
4 min read
Frontier AI models have outgrown current testing and benchmarking methods for hacking abilities. Federal agencies must establish a classified benchmarking process by Aug. 1 to assess model capabilities. Without new tests, policymakers and security teams cannot predict safe deployment.
Level
Hype check
Tap to vote and see what everyone thinks.
