AI Threat Output and Monitoring Incident Containment Act

AI-generated ?

Report an error

Introduced

In Committee

Passed One Chamber

Passed Both Chambers

Signed into Law

Referred to the House Committee on Science, Space, and Technology.

Creates a program to test advanced AI for nuclear risks, requiring big tech companies to participate.

This bill would create a new federal program to test advanced artificial intelligence (AI) for risks of nuclear incidents. It would require large AI companies to participate in these tests and follow safety rules. This changes the current situation where no such federal program or mandatory testing exists.

What Would Change

Today, there is no federal program requiring large artificial intelligence (AI) companies to test their advanced AI systems for the risk of nuclear incidents. If this bill becomes law, the Secretary of Energy would create such a program, requiring these companies to participate. Companies that do not comply could face civil penalties of up to $1 million per violation or per day for ongoing issues.

Key Provisions

Creates a new federal program to test advanced artificial intelligence (AI) for risks of nuclear incidents.
Requires large advanced AI companies to participate in this testing program.
Sets up testing methods, including 'red-teaming,' to find weaknesses in AI systems.
Prohibits large AI companies from refusing to participate, making false statements, or blocking the program.
Allows the Secretary of Energy to fine companies up to $1 million for each violation or for each day they do not comply.
Grants the Secretary of Energy the power to issue subpoenas to get AI systems and information for testing if companies do not cooperate.

View on Congress.gov

HR 9965 · 119th Congress · AI Summary by gemini-2.5-flash · 10/10

AI Threat Output and Monitoring Incident Containment Act

What Would Change

Key Provisions

Lobbying Activity