Collaboration on Adversarial Threats and Security Risks Act
- Sponsor
- Sen. Schiff, Adam B. [D-CA]
- Committees
- Judiciary Committee (primary)
- Last action
- Jul 23, 2026
Bottom line
The bill aims to enable critical collaboration among AI developers to mitigate severe AI security risks by providing a legal safe harbor from antitrust laws, while retaining government oversight to prevent abuse.
What it actually does
This bill creates a limited antitrust exemption for non-Federal entities that collaborate in good faith to address significant artificial intelligence (AI) security risks. It allows for sharing information or coordinating actions, such as delaying or limiting AI release, provided such activities are for the exclusive purpose of mitigating defined 'covered artificial intelligence security risks' and, in some cases, with prior notification to the Assistant Attorney General. The bill also grants the Attorney General authority to seek injunctions against actions that violate antitrust laws or increase AI security risks.
Proponents argue
Proponents argue that the rapid advancement of frontier AI models presents unprecedented national security and societal risks that no single entity can address alone. This bill is essential to foster necessary collaboration among developers and researchers, allowing them to share critical security information and coordinate risk mitigation strategies without fear of antitrust prosecution, thereby safeguarding national security and public safety.
Opponents contend
Opponents might contend that this antitrust exemption, even with safeguards, creates a dangerous precedent that could be exploited by dominant AI companies to collude, stifle competition, and slow innovation under the guise of security concerns. They may argue that the broad definitions of 'covered artificial intelligence security risk' could be misused to justify anti-competitive practices, ultimately harming consumers and smaller market entrants.
The bill is concise and can be read quickly, but its legal and policy implications are substantial and warrant careful consideration beyond a superficial review.
Section 1
Short Title
This section formally names the Act the 'Collaboration on Adversarial Threats and Security Risks Act.' This title succinctly communicates the bill's primary objective: to facilitate cooperation among entities to address threats and security risks associated with advanced artificial intelligence.
Supporters argue
Supporters would argue that a clear and descriptive short title aids in public understanding and legislative referencing, accurately reflecting the bill's critical purpose.
Critics contend
No specific opposition is typically raised against short title provisions, as they are standard legislative practice.
Tradeoffs
None apparent for this provision.
Section 2
Definitions of Key Terms
This section defines crucial terms, including 'antitrust laws' (by reference to existing statute), 'artificial intelligence,' 'assistance,' 'Assistant Attorney General,' 'covered artificial intelligence security purpose,' and 'covered artificial intelligence security risk.' The definition of 'covered artificial intelligence security risk' is particularly detailed, encompassing scenarios where AI could be weaponized by adversarial nations, facilitate the development of CBRN or offensive cyber weapons, disrupt critical infrastructure, lead to loss of operational control, or autonomously improve in a risky manner. It also defines 'unauthorized access' to AI.
Section 3
Antitrust Exemption for AI Security Collaboration
This section establishes that non-Federal entities will not violate antitrust laws when they provide or exchange information or assistance related to a covered AI security risk in good faith for an exclusive security purpose. It also exempts coordination or agreements to delay or limit AI release for such purposes, provided written notice detailing the risk and scope of restriction is submitted to the Assistant Attorney General. The exemption is limited to entities using information for security purposes with internal controls. It creates an affirmative defense for entities claiming the exemption, requiring proof of good faith and exclusive purpose. Crucially, it clarifies that the exemption does not permit price-fixing, market allocation, monopolization, boycotting, or exchanges of price/cost information.
Section 4
Attorney General's Authority for Injunctive Relief
This section grants the Attorney General the authority to seek an injunction in a court of competent jurisdiction against non-Federal entities to prohibit actions that violate antitrust laws, even if those entities claim the antitrust exemption. The exemption does not provide immunity if the entities fail to satisfy the burden of proof for good faith and exclusive purpose, or if the Attorney General demonstrates that the actions are reasonably likely to result in an overall increase in covered AI security risks.
Defines 'antitrust laws' by referencing section 102 of the Cybersecurity Information Sharing Act of 2015 (6 U.S.C. 1501). It also includes a rule of construction that the laws described in 6 U.S.C. 1501(2) are consistent to the extent they do not conflict.
Section 2(1)
Why it matters:Referencing existing definitions is standard legislative practice for consistency across federal law and to avoid redundant drafting. The rule of construction provides a necessary legal clarification for potential conflicts.
Case for: Using established definitions promotes legal consistency and efficiency, ensuring that the scope of antitrust laws covered by this bill aligns with existing federal frameworks.
Case against: Requires legal expertise and additional research to fully grasp the precise scope of the antitrust exemption, potentially obscuring the full implications for non-legal experts or the general public.
Estimated impact: Clarifies the specific legal framework under which the antitrust exemption operates, fundamentally shaping its application and enforcement.
Defines 'covered nation' by referencing section 4872(f)(2) of title 10, United States Code, in the context of AI being stolen, weaponized, or accessed by such nations or entities controlled by them.
Section 2(6)(A) and (F)(ii)(I)
Why it matters:This is standard legislative practice to ensure consistency with existing national security definitions and policies regarding adversarial nations, rather than creating a new, potentially conflicting list.
Case for: Ensures that the bill's national security focus aligns with established federal policy regarding foreign adversaries, promoting a unified approach to national defense and intelligence.
Case against: Requires external research to understand which specific nations are targeted, potentially making the bill's national security scope less immediately clear to a general audience or those unfamiliar with defense statutes.
Estimated impact: Directly influences the national security implications of the bill by identifying specific foreign adversaries whose actions related to AI constitute a 'covered artificial intelligence security risk.'
Stipulates that any information submitted to the Assistant Attorney General under subsection (a)(2) (regarding coordinated delays or limitations of AI release) shall be used solely for the purpose of Section 4 (injunctive relief), deemed voluntarily shared, and explicitly exempt from disclosure under the Freedom of Information Act (FOIA) (5 U.S.C. 552), to be withheld without discretion.
Section 3(e)
Why it matters:This provision is likely intended to encourage companies to share sensitive information about AI security risks and coordination efforts candidly, without fear of proprietary data being exposed or national security vulnerabilities being revealed to the public or adversaries.
Case for: Proponents argue that this confidentiality is essential to facilitate open and honest sharing of sensitive information among companies and with the government regarding AI security risks. Without such protection, companies might be hesitant to disclose critical data, undermining the bill's primary goal of fostering collaboration for security.
Case against: Critics might argue that this provision significantly reduces transparency and public oversight of potentially anti-competitive agreements made by powerful AI entities. It could make it harder for the public, journalists, or smaller competitors to scrutinize the justifications for coordinated delays or limitations in AI development, raising concerns about accountability.
Estimated impact: Creates a significant carve-out from public information laws for specific AI-related disclosures, affecting transparency, public accountability, and the balance between national security/proprietary interests and public access to information.