National security AI laws by state
AI Laws USA tracks 14 U.S. AI rules on national security AI across federal, state, county, and city government. Each entry links to its official source.
Federal national security AI rules (8)
-
Blocked / in litigation
Anthropic v. Dept. of War — D.C. Circuit upholds supply-chain-risk designation (2026)
United States · Anthropic PBC v. U.S. Department of War (D.C. Cir. Sept. 25, 2026)
On September 25, 2026, the U.S. Court of Appeals for the D.C. Circuit ruled 2-1 that the Defense Department could label AI company Anthropic a 'supply chain risk,' letting the Pentagon remove Anthropic's Claude models from its systems and bar their use for Defense Department work. Judge Gregory Katsas, joined by Judge Neomi Rao, wrote that the department reasonably feared Anthropic might design Claude to refuse national-security tasks; Judge Karen LeCraft Henderson dissented, arguing the 2018 supply-chain law was aimed at sabotage by foreign powers. The ruling is separate from, and points the other way from, a California federal court decision that found the Pentagon unlawfully retaliated against Anthropic. Anthropic said it disagrees and is considering further review.
-
In effect
Trump AI Innovation & Security EO (June 2026)
United States · Effective June 2, 2026 · E.O. (June 2, 2026) — Promoting Advanced Artificial Intelligence Innovation and Security
President Trump signed an executive order on June 2, 2026, directing frontier AI developers to voluntarily share new models with the federal government 30 days before public release for national-security review. The order also directs CISA to build an AI cybersecurity framework and tasks DOJ with prioritizing criminal enforcement of AI-enabled fraud. No binding requirements apply to private AI developers — the framework is voluntary.
-
Proposed / pending
Sanders-Casar Ban Artificial Superintelligence Act (2026)
United States · Ban Artificial Superintelligence Act (Sanders–Casar; unveiled Sept. 23, 2026)
Sen. Bernie Sanders and Rep. Greg Casar unveiled a bill on September 23, 2026 that would permanently ban artificial superintelligence — defined as AI that exceeds human cognitive ability or could plan and carry out the destruction or disempowerment of humanity. It would pause development of the most advanced AI systems until a new federal Department of Artificial Intelligence sets safety rules, after which advanced systems would need federal approval before deployment. Some violations could bring up to 20 years in prison. The AP reported the bill faces long odds in the Republican-controlled Congress.
-
Proposed / pending
Stop Flock Abuse Act (Hawley, 2026)
United States · Stop Flock Abuse Act (Sen. Hawley, announced Sept. 30, 2026; bill number pending)
Sen. Josh Hawley announced a federal bill, the Stop Flock Abuse Act, to set safeguards for AI-powered cameras such as Flock automated license plate readers (ALPRs). It would bar local governments from selling or sharing vehicle location data with non-government third parties, require a written approval and audit log for every search, require deletion of driver data after ten days except for active criminal investigations, require encryption and keep data inside the United States, ban hardware or software from foreign adversaries, and prohibit ALPR networks from using facial recognition. It follows a Senate Judiciary subcommittee hearing on Flock that Hawley chaired.
-
Proposed / pending
US S. 5601 DoD reporting requirements for large AI contractors
United States · S. 5601, 119th Cong. (2026)
A Senate bill that would require the Secretary of Defense to set reporting requirements and voluntary security guidance for large artificial intelligence contractors, focused on insider threats. It was introduced on Sept. 29, 2026 and sent to the Armed Services Committee. It has not passed either chamber.
-
Proposed / pending
H.R. 9917 AI Kill Switch Act
United States · H.R. 9917, 119th Congress
This House bill would amend the Homeland Security Act to require large AI companies to keep the technical ability to stop, suspend or shut down covered AI systems. Companies would have to report covered incidents to the Department of Homeland Security within 15 days and preserve model weights and telemetry when ordered. It would apply to entities earning at least $500 million a year from such systems that were built with more than $100 million worth of computing power.
-
Proposed / pending
H.R. 10502 AI use under FISA 702 report
United States · H.R. 10502, 119th Cong. (2026)
A House bill would require the Director of National Intelligence to tell Congress, within 180 days, how intelligence agencies use artificial intelligence to collect, search and share information gathered under Section 702 of the Foreign Intelligence Surveillance Act. The report would have to name the AI models used and assess safeguards for compliance with FISA and the Fourth Amendment.
-
Proposed / pending
H.R. 10670 China FIREWALL Act
United States · H.R. 10670, 119th Congress
This House bill would bar federal executive agencies from using or buying AI applications developed or provided by a 'foreign entity of concern,' including software built on such a company's foundation model. The Office of Management and Budget would have 60 days after enactment to issue removal standards. Exceptions would be allowed for law enforcement, national security and security research, with risk-mitigation documentation.
State national security AI rules (4)
-
In effect
California EO N-9-26 — accelerated AI oversight and kill-switch study
California · Effective September 18, 2026 · Cal. Exec. Order No. N-9-26 (Sept. 18, 2026)
California's governor signed an executive order on September 18, 2026 directing the Government Operations Agency to speed up how fast two already-enacted AI laws take effect — SB 813, which creates the framework for independent verification organizations that assess AI systems, and AB 1405, which creates the state registry and independence standards for AI auditors — and to convene national experts who have two months to recommend how to strengthen California's AI safety and security law. The order names four proposals the experts are asked to consider: putting a designated independent verification organization ONSITE inside frontier AI labs to run regular audits and evaluations; requiring that frontier companies' safety frameworks, transparency reports and risk assessments be VERIFIED against standards an independent verification organization deems adequate, rather than self-attested; advancing a 'kill switch' for frontier models whose effectiveness is itself verified on an ongoing basis by an independent verifier; and updating the definition of a critical safety incident to include LOSS-OF-CONTROL incidents. UNDERSTAND WHAT THIS ORDER IS. It imposes no new duty on any AI company today. It is an acceleration-and-recommendation instrument: it moves two existing statutes' timelines and it commissions a plan. The kill switch that led the headlines is a proposal the expert group is asked to consider, not a requirement now in force. The order's own framing is that the federal government has failed to act and that California will not wait — the substantive obligations, if they come, arrive through the recommendations and any legislation that follows.
-
Proposed / pending
PA HB 2800 frontier AI risk prevention
Pennsylvania · Pennsylvania HB 2800 — Frontier AI Risk Prevention: Standards for Frontier Developers on Critical Safety Incidents and Catastrophic Risks (Referred Sept. 23, 2026)
This Pennsylvania House bill would set standards for developers of the most advanced (frontier) AI models to address critical safety incidents and catastrophic risks. It gives duties to the Pennsylvania Emergency Management Agency and the Attorney General and includes penalties. It was referred to the House Communications and Technology Committee.
-
Proposed / pending
PA HR 637 urging federal frontier AI regulation
Pennsylvania · Pa. H.R. 637 (2025-2026 Sess.)
Pennsylvania House Resolution 637 urges Congress to pass federal legislation regulating frontier artificial intelligence models. It is non-binding. The House Communications & Technology Committee reported it out 26-0 on Oct. 7, 2026; the full House has not voted.
-
Proposed / pending
PA HB 2829 Advanced AI Accountability and Innovation Act
Pennsylvania · Pennsylvania HB 2829 — Pennsylvania Advanced Artificial Intelligence Accountability and Innovation Act: Frontier Developer Safety Frameworks and Transparency (Referred Oct. 5, 2026)
This Pennsylvania House bill would set standards and transparency rules for the safety frameworks of frontier AI developers. It gives duties to the Pennsylvania Emergency Management Agency and the Attorney General and includes penalties. It was referred to the House Communications and Technology Committee.
County national security AI rules (1)
-
In effect
Austin County TX data center and BESS moratorium (2026)
Austin County, TX · Effective July 27, 2026 · Austin County, Tex. — countywide moratorium on AI data centers and battery energy storage systems (adopted July 27, 2026)
Austin County Commissioners unanimously approved a countywide moratorium on artificial intelligence data centers and battery energy storage systems on July 27, 2026, becoming one of the first Texas counties to do so. The moratorium does not outright bar construction; it pauses approvals so the county can evaluate what these developments do. What sets it apart from almost every other pause in this dataset is what the county said it will examine and what it demands of developers in the meantime. The evaluation covers compliance with health, fire and safety requirements, CYBERSECURITY SAFEGUARDS, and enforcement mechanisms that can actually be applied to these projects in the unincorporated parts of the county — including the possible designation of a dedicated energy development zone. And the moratorium REQUIRES prospective developers to submit an equipment manufacturer audit to the county: a full list of every piece of equipment and every component proposed for use in the project. That is a supply-chain disclosure requirement, and no other local measure indexed here asks for one. It gives a rural county visibility into who made the hardware going into a facility on its land, which is a fire-safety question for battery storage and a security question for computing. Austin County was the second Texas county to adopt a data center moratorium after Hill County; San Marcos was the first Texas city, on June 30, 2026.
City / local national security AI rules (1)
-
In effect
Ypsilanti Township MI electric utility infrastructure moratorium (2026)
Ypsilanti Township, MI · Effective August 18, 2026 · Charter Township of Ypsilanti, Mich. — six-month land-use moratorium on major electric utility facilities (adopted August 18, 2026)
Ypsilanti Township's Board of Trustees unanimously approved a land-use moratorium on August 18, 2026 — and the thing it regulates is not data centers. It places a SIX-MONTH PAUSE on land use for the establishment or substantial expansion of a MAJOR ELECTRIC UTILITY FACILITY. That is a genuinely different instrument from everything else in this dataset, and it is the most transferable idea recovered in this triage. A hyperscale computing facility cannot run without dedicated high-capacity power, which in practice means a new on-site substation. A township that regulates the SUBSTATION reaches the project without ever having to define, classify or prohibit a 'data center' — and without the definitional fights that have forced other jurisdictions to amend their codes. The target here is specific. The University of Michigan had announced days earlier that it had selected a preferred location for what it describes as a computing research facility, on a Textile Road plot the university confirmed purchasing in March. The project includes a U-M academic research facility AND A CLASSIFIED RESEARCH FACILITY FOR LOS ALAMOS NATIONAL LABORATORY's work — and an electrical substation, which U-M said it would build with DTE Energy as a dedicated on-site facility. Township Supervisor Brenda Stumbo said the pause allows the township to evaluate and adopt ordinances regulating the kind of infrastructure required to power projects like it. The township is therefore asserting land-use authority over the power supply to a federal national-laboratory classified facility, which is an unusual posture worth watching.