HR 10981: American AI Security Act

HR 10981 in plain English: This bill would establish security requirements related to artificial intelligence, with enforcement authority given to the Secretary to impose civil penalties for violations. It has been referred to the House Committees on Energy and Commerce and Intelligence for further consideration.

Stated purpose

This bill directs the Secretary of Commerce to create regulations for certain large artificial intelligence models, requiring developers to submit those models to the National Security Agency for review before releasing them publicly, in order to protect national security.

Key points

• Grants the Secretary authority to issue civil penalties of $500,000 per day for non-compliance.
• Referred to both the Energy and Commerce Committee and the Permanent Select Committee on Intelligence.

Arguments supporters make

• Powerful AI models could help bad actors create cyberattacks or weapons of mass destruction, so a government review before public release is a reasonable safety check.
• Requiring developers to share safety test results and deployment plans with the NSA closes a gap where dangerous AI capabilities might go undetected until it is too late.
• Annual reviews of the regulations ensure the rules keep pace with rapidly changing AI technology rather than becoming outdated.

Arguments opponents make

• Requiring government approval before releasing an AI model could slow innovation, put American companies at a disadvantage against foreign competitors who face no such delays, and give bureaucrats too much control over private technology.
• Giving the NSA full access to a company's AI model, training methods, and business plans raises serious concerns about government overreach, trade secret exposure, and potential misuse of sensitive commercial information.
• The 30-day review window may be too short for a thorough evaluation but still long enough to cause harmful delays, leaving the process neither safe enough nor fast enough to work well in practice.

Tradeoffs

Screening AI models before release may reduce the risk of catastrophic misuse, but it also adds government control and delays to a fast-moving industry, creating tension between national security goals and the speed and freedom that drive technological innovation.

Current status in Congress: In committee.

NewsClear — neutral news & congressional tracking · Bill of the Week