Overview
- Sen. Maria Cantwell released the six-point framework Wednesday to form the basis for federal legislation governing so-called frontier AI models that pose high-risk threats.
- The plan would task the National Institute of Standards and Technology with writing risk-based safety standards for systems that could enable cyberattacks, biological or nuclear misuse, loss of human control, or autonomous agents escaping test environments.
- It would require continuous independent testing and pre-release audits, public disclosure of serious failures, whistleblower protections, and special rules for open-source models to limit misuse while preserving research paths.
- The proposal won cautious praise from some industry leaders but arrives alongside a voluntary White House-industry safety pact and new California laws that create a fragmented patchwork of rules and voluntary commitments.
- Passage is uncertain because Senate negotiations are stalled and committee control matters to turning the framework into law, but the plan signals where federal enforcement, worker and child protections, and international incident coordination could be focused next.