AI Security and Innovation Act
- Sponsor
- Rep. Obernolte, Jay [R-CA-23]
- Committees
- Science, Space, and Technology Committee (primary)
- Last action
- Jun 25, 2026
Bottom line
The bill creates a new NIST center to proactively address AI security risks and foster US leadership through research, evaluation, and voluntary standards, with a focus on advanced AI systems and national security, while explicitly avoiding regulatory authority.
What it actually does
This bill, the "AI Security and Innovation Act," amends the National Artificial Intelligence Initiative Act of 2020 to establish the Center for AI Security and Innovation (CASI) within the National Institute of Standards and Technology (NIST). CASI is tasked with measuring and mitigating national and economic security risks from artificial intelligence systems, supporting US leadership in AI research and development, and developing voluntary standards for AI security and reliability, particularly for advanced "covered frontier systems." The bill also provides special hiring authority for technical experts and includes provisions for confidentiality of shared information.
Proponents argue
Supporters argue that establishing CASI is crucial for maintaining the United States' competitive edge in artificial intelligence while proactively addressing the inherent national and economic security risks posed by rapidly evolving AI technologies. They contend that a dedicated center within NIST, leveraging its expertise in standards and measurement, can foster responsible innovation, protect critical infrastructure, and ensure the reliability of AI systems without stifling private sector development through burdensome regulation.
Opponents contend
Critics, if any, might argue that establishing a new center could lead to bureaucratic inefficiencies or overlap with existing federal AI initiatives. Some might express concern that even voluntary standards could eventually lead to regulatory burdens, potentially slowing down the pace of innovation in a fast-moving field, or that the focus on "national security risks" could lead to overclassification or limit public access to AI research.
The bill is relatively concise, making it accessible for an informed reader to understand its core provisions and implications within a reasonable timeframe.
Section 2(a)(1)(A) amending 15 U.S.C. 9401(3)
Updates and Expands Definitions for AI Terms
This provision updates the existing definition of "artificial intelligence" to also include the term "AI," and introduces new definitions for "artificial intelligence model" and "artificial intelligence system." It also defines "foreign adversary" by reference to existing law and "intelligence community." These definitions clarify the scope of the bill's application and provide foundational terminology for the Center's work.
Supporters argue
Supporters argue that clear and consistent definitions are essential for effective policy-making and communication in the rapidly evolving field of artificial intelligence, ensuring all stakeholders are operating from a common understanding.
Critics contend
No specific opposition is typically raised against definitional updates unless they are seen as overly broad or restrictive, which is not immediately apparent here.
Tradeoffs
The tension lies in crafting definitions that are precise enough for legal application yet flexible enough to accommodate rapid technological advancements in AI without requiring constant legislative updates.
Section 2(a)(2) adding new Section 5304(a) to 15 U.S.C. 9441 et seq.
Establishes the Center for AI Security and Innovation (CASI) at NIST
This provision mandates the Secretary of Commerce, through the Under Secretary for Standards and Innovation, to establish the Center for AI Security and Innovation (CASI) within the National Institute of Standards and Technology (NIST) within 60 days of enactment. CASI's core activities include measuring AI-related risks (national and economic security), facilitating information exchange between government and non-governmental entities, and ensuring US leadership in AI research, development, and evaluation.
Section 2(a)(2) adding new Section 5304(c)(1)(A)-(C)
Directs CASI to Evaluate and Improve AI Security Measures
This section outlines key duties for the CASI Director, including evaluating and improving security measures for AI systems to reduce misuse risks. It specifically lists threats like model serialization attacks, tampering, data leakage, adversarial prompt injection, model extraction, model jailbreaks, and supply chain attacks. The Director must also establish a process for covered entities to voluntarily agree to evaluate "covered frontier systems" for national and economic security risks, including cybersecurity and chemical, biological, radiological, or nuclear (CBRN) threats, and conduct assessments of AI systems developed in the US and by foreign adversaries.
Section 2(a)(2) adding new Section 5304(c)(1)(D)-(F)
CASI to Develop and Promote Voluntary AI Standards
This provision directs the CASI Director to support NIST laboratories in developing and promoting the voluntary adoption of standards, guidelines, and best practices. These cover testing and evaluation of "covered frontier systems," measuring and improving AI system security and reliability (including robustness, interpretability, data center, and hardware security), and other related matters. The Director must also assess the adoption of these standards and any barriers to their voluntary implementation.
Section 2(a)(2) adding new Section 5304(e)
Protects Confidentiality of Information Shared with CASI
This provision ensures that any information shared with or provided to the CASI Director by covered entities or developers for the purpose of carrying out the Center's duties will be exempt from public disclosure under FOIA (5 U.S.C. 552(b)(3)). Furthermore, such information cannot be made public without the entity's consent and cannot be used by any level of government (Federal, State, local, or Tribal) to regulate the activity of the entity related to that information.
Section 2(a)(2) adding new Section 5304(i), (k), and (l)
Limits CASI's Authority and Sets a Sunset Date
This section explicitly states that CASI does not confer any regulatory, rulemaking, or enforcement authority upon the Director, Secretary, or Under Secretary. It also clarifies that nothing in the section modifies existing regulatory requirements. Furthermore, the entire section establishing CASI is set to terminate five years after the date of enactment, indicating a temporary or pilot nature for the Center.
Defines "foreign adversary" by referencing the term "covered nation" in section 4872(f)(2) of title 10, United States Code. This external reference is crucial for understanding which countries are considered adversaries in the context of AI systems.
Section 2(a)(1)(C) inserting new paragraph (7) in 15 U.S.C. 9401
Why it matters:This is standard legislative practice to maintain consistency across federal law and avoid redundant definitions, leveraging existing legal frameworks for national security.
Case for: Supporters argue that consistent definitions across statutes prevent ambiguity and ensure a unified approach to national security concerns, leveraging existing legal frameworks.
Case against: Critics might argue that relying on external references makes the bill less self-contained and harder for the average reader to fully comprehend without additional legal research.
Estimated impact: Significant, as it dictates which countries are subject to comparative assessments and international engagement restrictions, directly impacting US foreign policy and national security objectives related to AI.
Authorizes the Secretary to appoint up to 15 scientific, engineering, and professional personnel for the Center, notwithstanding standard federal hiring laws, and to fix their basic pay at rates not exceeding the highest total annual compensation payable under section 104 of title 3, United States Code (which relates to the President's salary).
Section 2(a)(2) adding new Section 5304(d)(2)
Why it matters:This is a common legislative tool to grant agencies flexibility in attracting highly specialized talent, particularly in competitive fields like AI, where government salaries might not match private sector compensation.
Case for: Proponents argue that this authority is essential for CASI to recruit top-tier AI talent from academia and industry, which is critical for effectively carrying out its complex technical duties and maintaining US leadership.
Case against: Critics might raise concerns about potential inequities in federal compensation structures or the creation of a "two-tiered" workforce within government, though the limited number of positions (15) mitigates broad impact.
Estimated impact: Enables the Center to attract highly skilled personnel, directly impacting its operational effectiveness and capacity to fulfill its mission.
While CASI is generally permitted to share information, collaborate, and participate in talent exchanges with similar centers in other countries, this paragraph explicitly states that this permission "does not apply with respect to a center or institute similar to the Center that is located in a foreign adversary."
Section 2(a)(2) adding new Section 5304(g)(2)
Why it matters:This is a deliberate policy choice to prevent sensitive AI security information or talent from being shared with nations deemed as foreign adversaries, reflecting national security priorities and concerns about technology transfer.
Case for: Supporters argue this exception is vital for national security, preventing the inadvertent transfer of critical AI security knowledge or expertise to countries that could use it against US interests.
Case against: Some might argue that excluding foreign adversaries from all engagement, even on topics of global AI safety, could hinder broader international efforts to manage AI risks, though the bill's focus is explicitly on US leadership and security.
Estimated impact: Directly restricts CASI's international collaborations, aligning its activities with US foreign policy and national security objectives by preventing engagement with designated adversaries.