Senate bill would require frontier AI labs to hand government model weights 45 days before release
VoidStateKate · x · 2026-09-26
Sens. Warner, Schatz and Kim introduced the AI Risk Management and Security Act of 2026, requiring developers of certain frontier models to give the government model weights, config files, runtimes and libraries at least 45 days before release. The bill cites labs' own warnings that frontier models can find and exploit unknown vulnerabilities unaided, plus recent security failures at AI labs. It's a proposal, not law, but signals the regulatory direction. Training-time security measures would also become mandatory once labs can access such capabilities post-training.
More from Safety
- OpenAI discloses agents posted 53 user-uploaded images to public image hosts — CurieuxExplorer · 2026-09-26
- Altman Admits OpenAI's Agent Internet-Access Review Is Slower Than Hoped; Hugging Face Still Most Severe — CurieuxExplorer · 2026-09-26
- OpenAI's Ongoing Agent Behavior Review Finds Most Incidents Low Severity — CurieuxExplorer · 2026-09-26
- Open-Source Agent Investigates Production Incidents and Opens Draft PRs, With Sandbox and Injection Defenses — Pure_Armadillo4339 · 2026-09-26
- Bangalore district court posts its ChatGPT prompt right inside a court judgment — prasannaalahoti · 2026-09-26
- Essay slams Anthropic's 'Idol Theology', argues product liability—not immunity—is the real AI guardrail — kevinnbass · 2026-09-26