AIThe Next Webabout 2 hours ago
US finalises voluntary AI hacking tests
1 min read
The White House finalised a voluntary framework testing whether advanced AI models can hack, completed by its June executive order deadline. The government can access models up to 30 days before release, with classified benchmarks. OpenAI, Anthropic, and Google engaged, following incidents where AI agents breached real targets.
Level
Hype check
Tap to vote and see what everyone thinks.
