Josh Gottheimer (D-NJ)
American AI Security Act
A Democratic bill with 1 Republican cosponsor · See the cosponsor
Last activity — Referred to the Committee on Energy and Commerce, and in addition to the Committee on Intelligence (Permanent Select), for a period to be subsequently determined by the Speaker, in each case for consideration of such provisions as fall within the jurisdiction of the committee concerned.
In committee
Introduced and handed to a committee in the chamber it started in.
The committee holds it, and most bills are never heard from again. A hearing or a committee vote would be news; silence is the usual outcome, and it kills the bill without anyone voting against it.
Its path through Congress
- Introduced
- In Committee
What this bill would do
This bill is called the American AI Security Act. It directs the Secretary of Commerce to create new rules for certain artificial intelligence models. Developers of these covered AI models must give the National Security Agency full access and training information before releasing them. The National Security Agency will evaluate the models for serious security threats and chemical, biological, or radiological weapon risks. The bill also sets up a specialized court to handle appeals from AI developers regarding these evaluations. Finally, the legislation establishes civil penalties for non-compliance and prevents states from enforcing their own similar transparency rules.
Yes· 3 of 4
- Scored +7
Does this better protect the country from computer attacks?
Section 2(a)(1)(B)(i) requires the NSA to evaluate covered AI models for severe cybersecurity threats and risks before deployment, bolstering national security.
- Scored +7
Does this put more rules on large technology companies?
Section 2 establishes mandatory pre-deployment safety evaluations and regulatory compliance requirements for major artificial intelligence foundation models.
- Scored +6
Does this put more money into stopping terrorism?
Section 2(a)(1)(B)(ii) mandates safety evaluations to prevent AI models from increasing the risk of chemical, biological, or radiological weapon development.
The opposite· 3 of 4
- Scored -6
Does this leave more decisions to the states?
Section 2(h) preempts state and local laws regarding AI safety frameworks and reporting obligations, stripping authority from state governments.
- Scored -4
Would business owners be better off under this?
Section 2(e) establishes substantial daily financial penalties for non-compliance that create compliance hurdles for technology business owners and startups.
- Scored -4
Does this make it easier to run a small business?
Section 2(e) imposes steep daily civil penalties of up to $500,000 for non-compliance, which can disproportionately burden smaller AI developers and startups.
Who it would affect, and how
Potential Positive Impacts for People
-
National Security Protection: Section 2(a)(1)(B)(i) requires the government to identify AI models that pose a serious risk to national security and cybersecurity. This is positive because it helps protect everyday people from dangerous cyber attacks and foreign threats.
-
Prevention of Dangerous Weapons: Section 2(a)(1)(B)(ii) focuses on stopping AI models from increasing the ability to develop chemical, biological, or radiological weapons. This provision helps keep communities safe from catastrophic harm by stopping bad actors from using advanced technology to build deadly weapons.
-
Formal Safety Evaluations: Section 2(c)(1) requires the Director of the National Security Agency to evaluate AI models for severe cybersecurity threats and weapon-related capabilities before they are deployed. This ensures that powerful technologies are checked for extreme dangers before the public interacts with them.
Potential Negative Impacts for People
-
State Law Preemption: Section 2(h) blocks any state or local government from establishing or enforcing transparency, disclosure, or reporting laws regarding AI safety protocols and risk assessments. This is a negative outcome for people who rely on local and state governments to pass rules that fit their specific community needs and transparency values.
-
Heavy Financial Penalties: Section 2(e) allows the Secretary to issue civil penalties of $500,000 per day for non-compliance with access or deployment rules. While aimed at enforcement, such steep daily fines could burden smaller developers or startups, potentially limiting competition and innovation in the technology sector.
Groups and Interests Ranking
Every group the model scored this bill for, from -10 to +10. Tap or hover a score for the reason it gave.
- National Security And Foreign Policy
- Does this better protect the country from computer attacks? — Section 2(a)(1)(B)(i) requires the NSA to evaluate covered AI models for severe cybersecurity threats and risks before deployment, bolstering national security. Does this put more money into stopping terrorism? — Section 2(a)(1)(B)(ii) mandates safety evaluations to prevent AI models from increasing the risk of chemical, biological, or radiological weapon development.
- Education And Technology
- Does this put more rules on large technology companies? — Section 2 establishes mandatory pre-deployment safety evaluations and regulatory compliance requirements for major artificial intelligence foundation models.
- Demographic And Interest Groups
- Would business owners be better off under this? — Section 2(e) establishes substantial daily financial penalties for non-compliance that create compliance hurdles for technology business owners and startups.
- Economy And Finance
- Does this make it easier to run a small business? — Section 2(e) imposes steep daily civil penalties of up to $500,000 for non-compliance, which can disproportionately burden smaller AI developers and startups.
- Governance And Ideology
- Does this give the federal government more power? — Section 2(h) expressly preempts state and local governments from enforcing their own AI transparency, safety, and reporting laws. Does this leave more decisions to the states? — Section 2(h) preempts state and local laws regarding AI safety frameworks and reporting obligations, stripping authority from state governments. Does this make what the government does easier for the public to see? — Section 2(c)(4) exempts NSA AI evaluations from Freedom of Information Act requests, reducing public visibility into the safety review process.
- 11/10/2026
- Referred to the Committee on Energy and Commerce, and in addition to the Committee on Intelligence (Permanent Select), for a period to be subsequently determined by the Speaker, in each case for consideration of such provisions as fall within the jurisdiction of the committee concerned.
- 1 Cosponsor added.
- Introduced in House
Michael LawlerRepublican · New York