Advanced AI Security Readiness Act
- Bill Number
- H.R. 3919
- Origin Chamber
- House
- Congress
- 119th Congress, Session 1
- Policy Area
- Armed Forces and National Security
- Status
- Introduced
- Latest Action
- 2025-06-11: Referred to the House Committee on Intelligence (Permanent Select).
- Last Updated
- 2026-04-17T16:28:01Z
AI-Generated Summary
Purpose
The Advanced AI Security Readiness Act (H.R. 3919) aims to protect advanced artificial intelligence (AI) technologies from theft by foreign adversaries or other high-capability threats. It directs the National Security Agency (NSA) to create a comprehensive guide—called the "AI Security Playbook"—to identify risks and develop defenses, ensuring the U.S. maintains a strategic edge in AI development without imposing new regulations.
Key Provisions
- Development of the AI Security Playbook: The NSA Director, through the Artificial Intelligence Security Center (or a successor office), must create strategies to safeguard "covered AI technologies" (advanced AI systems with critical capabilities in areas like chemical/biological threats, cyber operations, or self-improving AI that could endanger national security if stolen).
- Core Elements of the Playbook:
- Identification of unique vulnerabilities in AI data centers and among developers, such as cybersecurity risks not common in standard IT systems.
- Listing of sensitive components (e.g., AI models, training data, or engineering insights) that, if accessed by threats, could accelerate their AI progress.
- Methods to detect, prevent, and respond to cyber threats targeting AI.
- Assessment of security levels for highly advanced AI that might require U.S. government involvement.
- Hypothetical plans for building secure AI in a government-controlled environment, covering cybersecurity, insider threat mitigation (e.g., background checks), access controls, and anti-espionage measures.
- Format and Dissemination: The Playbook includes a classified section with detailed intelligence and an unclassified version with practical guidelines for private sector use.
- Engagement Process: The NSA must consult AI developers, researchers, and experts through reviews, interviews, discussions, facility visits, and collaboration with a federally funded research center. These activities are exempt from the Federal Advisory Committee Act (FACA), which governs formal advisory groups.
- Reporting Requirements:
- An initial progress report to congressional intelligence committees within 90 days of enactment.
- A final report within 270 days, including unclassified and publicly available versions, with an optional classified annex.
- Definitions: Key terms include "covered AI technologies" (AI posing grave risks if stolen), "technology theft" (unauthorized acquisition via hacks, insiders, etc.), and "threat actors" (nation-states or well-resourced groups).
Significant Changes to Existing Law
This bill introduces new mandates for the NSA without amending prior laws. It builds on existing AI definitions from the 2019 National Defense Authorization Act but adds specific requirements for a playbook focused on theft prevention. Notably, it clarifies that the government cannot use the Playbook to enforce regulations or take regulatory actions, preserving voluntary compliance.
Potential Impacts
- Government Agencies: The NSA will need to allocate resources for playbook development, consultations, and reporting, potentially increasing coordination with intelligence and defense entities. It could lead to more secure government-led AI projects without mandating oversight.
- Citizens and Private Sector: AI companies and researchers gain access to non-binding best practices to protect their innovations, reducing risks of intellectual property theft. Citizens may indirectly benefit from stronger national security against AI-enabled threats like cyberattacks.
- International Relations: By focusing on threats from nation-states, the bill could heighten U.S. efforts to counter AI advancements by rivals (e.g., China or Russia), potentially straining relations if perceived as overly defensive. It promotes U.S. leadership in secure AI without direct international mandates.
Main Stakeholders Affected
- National Security Agency (NSA): Primary implementer, responsible for playbook creation and engagement.
- AI Developers and Researchers: Private sector entities involved in advanced AI, who provide input and receive security guidelines.
- Congressional Intelligence Committees: House and Senate Permanent/Select Committees on Intelligence, which receive reports and oversee progress.
- U.S. Government Agencies: Broader involvement in potential secure AI initiatives, including defense and research bodies.
- Threat Actors: Indirectly targeted, as the bill aims to deter nation-states and other actors from stealing AI tech.
Notable Legal, Constitutional, or Political Implications
- Legal: The bill avoids creating enforceable regulations, limiting it to advisory strategies and respecting private sector autonomy. Exemption from FACA streamlines consultations without bureaucratic hurdles.
- Constitutional: Aligns with Congress's oversight of national security (Article I powers) and executive branch intelligence roles, without infringing on free speech or private innovation rights.
- Political: Sponsored bipartisanship (e.g., by Reps. LaHood, Moolenaar, Gottheimer, Krishnamoorthi) signals broad consensus on AI as a national security priority. It emphasizes voluntary measures to balance innovation with protection, potentially influencing future AI policy debates amid global tech competition.
This summary was generated by AI and may contain inaccuracies. Refer to the official source document for the authoritative text.
Sponsor
Cosponsors (5)
Rep. Moolenaar, John R. [R-MI-2], Rep. Gottheimer, Josh [D-NJ-5], Rep. Krishnamoorthi, Raja [D-IL-8], Rep. Vindman, Eugene Simon [D-VA-7], Rep. Harder, Josh [D-CA-9]
Recent Actions
- 2025-06-11: Referred to the House Committee on Intelligence (Permanent Select).
- 2025-06-11: Introduced in House
- 2025-06-11: Introduced in House
Bill Versions
- Advanced AI Security Readiness Act — issued 2025-06-11 — PDF (7 pages)