U.S. Senate S.1792 "AI Whistleblower Protection Act" — shielding those who report AI security flaws or violations from retaliation

1AI safety and whistleblowers

S.1792, the "AI Whistleblower Protection Act," was introduced by Sen. Chuck Grassley (R-IA) with Sens. Coons, Blackburn, Klobuchar, Hawley, and Schatz on May 15, 2025 — a strongly bipartisan group — and referred to the Senate HELP Committee.

2How whistleblowers would be protected

The bill prohibits employment discrimination against employees (whistleblowers) who report AI security vulnerabilities or AI violations.

It defines an "AI security vulnerability" as a failure or lapse in security that could allow emerging AI to be acquired (including by a foreign entity) through theft or other means; and an "AI violation" as (A) any violation of federal law related to the development, deployment, or use of AI, or (B) failure to appropriately respond to a substantial and specific danger that AI development, deployment, or use may pose to public safety, public health, or national security.

"Artificial intelligence" is defined broadly, including machine learning and neural networks.

3Regulate from outside, or have it reported from inside

There are two routes to assuring the safety of AI: constraining it from outside by regulation, and encouraging reporting from inside. This bill takes the second.

The definitions are broad. An AI security vulnerability is a security failure or flaw through which emerging AI technology could be obtained by a third party, including a foreign actor, by theft or other means.

An AI violation covers violations of federal law relating to the development, deployment or use of AI, and failure to respond appropriately to a substantial and specific danger to public safety, public health or national security. Senior members of both parties among the sponsors is a further feature.