White House Keeps AI Cybersecurity Framework Secret, Raising Concerns

emmanuelvivier · x · 2026-08-10

According to WIRED, the US administration has finalized a plan to address cybersecurity risks posed by advanced AI models, but is deliberately keeping the details under wraps.

The White House invited leading AI companies like OpenAI, Anthropic, and Google to share an overview of the new framework. Developers can voluntarily submit models to the federal government up to 30 days before public release. The government will then vet their cyber capabilities using a classified benchmarking system. However, testing criteria and scope remain undisclosed, with open models reportedly excluded. Critics worry this secretive process could entrench large companies and disadvantage smaller startups.

Original post →

More from Safety

Safety channel →