NEWS: Sens. Schiff and Banks, Reps. Latta and Whitesides Introduce Bipartisan Bill to Combat AI Distillation and Other Attacks to National Security

Washington, D.C. - Today, U.S. Senators Adam Schiff (D-Calif.) and Jim Banks (R-Ind.), and Representatives Bob Latta (R-Ohio-5) and George Whitesides (D-Calif.-27) are introducing new bipartisan legislation to enhance the ability for AI labs and security researchers to combat distillation attacks from foreign adversaries and other frontier AI risks that pose a national security risk to the U.S.

The Collaboration on Adversarial Threats and Security Risks Act would allow for increased information sharing between industry and stakeholder groups to share information about AI security threats and coordinate their responses with other model developers. It includes strict guardrails, including protection against anti-competitive behavior, and permits the Attorney General to seek an injunction against non-Federal entities that abuse the exemption.

During distillation attacks, foreign labs systematically harvest data from U.S. AI models to produce their own advanced AI models at a fraction of the cost, posing an enormous threat to continued U.S. leadership and investment in developing the most powerful and efficient AI models and staying ahead of foreign competition. With the rise of AI, foreign labs have increasingly attacked proprietary models with millions of queries and use the resulting data to train rival, foreign-based systems, essentially reverse-engineering American technology at a lower cost and with less strain on valuable computational resources.

"We cannot allow China, or any foreign adversary, to steal American technological advancements in AI, and then weaponize those advances to undermine our own national security," said Senator Schiff. "The solution we are proposing would protect against AI distillation attacks and other security risks by allowing increased information-sharing among U.S. AI labs to counter the threat of foreign theft. As new AI innovations are developing faster than ever, it's critical that Republicans and Democrats continue to work together to address potential misuse of AI development by hostile actors at home and abroad."

"The AI race is one that America can't afford to lose. Communist China is working every day to steal our technology and close the gap with the United States," said Senator Banks. "American AI companies must be able to collaboratively defend against our adversaries. This bill gives them the tools to protect America's technological edge and helps ensure we maintain our competitive edge."

"Protecting the American people means ensuring the systems they rely on every day, including our hospitals, water systems, and electric grid, remain safe, secure, and reliable," said Congressman Latta. "Our adversaries are constantly looking for new ways to target our critical infrastructure, and we must make sure we are prepared to respond. This bipartisan legislation, which I am proud to lead alongside Senators Schiff and Banks and Rep. Whitesides will help strengthen collaboration between American companies and ensure that when a threat is identified, we can quickly share information and take action to protect our communities. Our adversaries are working together to undermine us, and we must work together to defend America."

"As the Representative for a district at the heart of California's aerospace and technology industry, I've seen firsthand how vital American innovation is to our national security. But as AI models grow exponentially more capable, the risk of these systems being exploited or weaponized poses a direct threat to public safety," said Congressman Whitesides. "By authorizing threat intelligence sharing to defend against hostile foreign attacks, this bipartisan bill prevents our technology from being turned against us. Safeguarding our technological edge isn't a partisan issue, and I'm proud to help lead this effort to keep America safe."

Specifically, the Collaboration on Adversarial Threats and Security Risks Act would create a statutory exemption that allows non-Federal entities to:

• Share information related to a covered AI security risk in good faith for the exclusive purpose of preventing, investigating, or mitigating an AI security risk; or
• Coordinate with the exclusive purpose of reducing AI security risks by delaying or limiting the deployment of high-risk AI models, with required written notice prior to taking such actions.

The covered coordination efforts include information-sharing and agreements made through coordinating bodies such as the NSA's AI Security Center, NIST's CAISI, the Frontier Model Forum, or other industrial consortia. The legislation tracks the Cybersecurity Information Sharing Act, which has successfully enabled the sharing of cyber threat and vulnerability information between and among government and nongovernmental organizations for a decade.

In addition to Schiff, Banks, Latta, and Whitesides, the legislation is cosponsored by U.S. Representatives Pat Harrigan (R-N.C.-10), Darrell Issa (R-Calif.-48), Ted Lieu (D-Calif.-36), Mariannette Miller-Meeks (R-Iowa-01), Nathaniel Moran (R-Texas-01), and Jay Obernolte (R-Calif.-23).

The Collaboration on Adversarial Threats and Security Risks Act is endorsed by the AI Policy Network (AIPN), Alliance for a Better Future (ABF), Alliance for Secure AI, America First Policy Institute (AFPI), Association for Competitive Technology (ACT), Center for AI Safety (CAIS) Action Fund, Encode AI, Foundation for American Innovation (FAI) Action, Google, Health Information Sharing and Analysis Center (Health-ISAC), Law Reform Institute, SeedAI, and Software & Information Industry Association (SIIA).

"The worst AI security threats - a foreign adversary stealing a frontier model, a serious loss of control - won't be contained to a single company. Developers need to be able to share threat information and coordinate, and this bill lets them do that. When companies can coordinate on threats without legal risk, they lose the excuse for staying silent and the public is safer for it," said Sunny Gandhi, Co-Executive Director, Encode AI.

"We urgently need legislation that lets frontier AI developers collaborate on national security risks, such as adversarial distillation, without being deterred by uncertainties regarding antitrust law. The Collaboration on Adversarial Threats and AI Security Risks Act would provide targeted protections paired with appropriate guardrails against misuse," said Tim Schnabel, President, Law Reform Institute.

"Building on the proven foundation of CISA 2015, The Collaboration on Adversarial Threats and Security Risks Act brings a strengthened threat-sharing framework into the AI era. Protecting critical infrastructure and cutting-edge AI systems requires real-time coordination against sophisticated security threats. This bill strikes a vital balance: empowering tech leaders to collaborate on cyber, CBRN, loss-of-control, and foreign adversarial risks while maintaining strong antitrust protections," said Paul Lekas, Executive VP for Global Public Policy & Government Affairs, Software & Information Industry Association (SIIA).

"AI developers are confronting increasingly sophisticated threats involving model theft, unauthorized access, critical infrastructure, and other serious security risks. The Collaboration on Adversarial Threats and Security Risks Act would extend existing protections to information sharing and coordination around AI-specific threats, promoting real-time situational awareness among frontier labs and other AI developers. Importantly, the bill includes safeguards to ensure this protection is limited to good-faith security collaboration and does not excuse anticompetitive conduct," said Morgan Reed, President, Association for Competitive Technology (ACT).

"The benefits of the CATS Act extend far beyond AI tech companies. By removing antitrust barriers to sharing critical information about adversarial AI threats, this legislation directly protects downstream critical infrastructure. For Health-ISAC and the health sector, this means safeguarding national public health and patient safety from emerging security risks and ultimately keeping Americans safer," said Denise Anderson, President & CEO, Health-ISAC.

"Making America AI-ready means making institutions more resilient and equipped to respond to AI security risks. Giving companies a clear, lawful path to share threat information raises the security floor for the whole country while advancing American leadership in AI. Making America AI-ready means making institutions more resilient and equipped to respond to AI security risks. Giving companies a clear, lawful path to share threat information raises the security floor for the whole country while advancing American leadership in AI," said Joshua New, Director of Policy, SeedAI.

"This bipartisan bill gives the AI industry clear rules of the road to share threat intelligence and collaborate on much-needed standards and benchmarks. We appreciate the leadership of Senators Schiff and Banks and Representatives Whitesides and Latta in working to ensure the continued responsible development of frontier models," said John Connell, VP of Government Affairs and Public Policy, US & Canada, Google.

"I commend Senators Schiff and Banks for their bipartisan legislation. The Collaboration on Adversarial Threats and Strategic Risks Act improves the speed at which national security risks arising from adversarial espionage and autonomous improvement of artificial intelligence models can be identified and addressed," said Mark Beall, President AI Policy Network.

"Advanced AI systems could be exploited by foreign adversaries, used to develop biological or cyber weapons, threaten critical infrastructure, or operate beyond meaningful human control. The Collaboration on Adversarial Threats and Security Risks Act is a commonsense step that will help America's leading AI companies work together to identify and reduce these risks before they become an emergency," said Brendan Steinhauser, CEO, The Alliance for Secure AI.

"As advanced AI systems increasingly become prime targets for espionage and theft by bad actors and global competitors, AI companies need a secure way to share threat information and coordinate defenses without risking antitrust violations. The Collaboration on Adversarial Threats and Security Risks Act would create a narrow exemption for companies to rely on, while keeping safeguards against price fixing, market allocation, and other anti-competitive behavior fully intact. By strengthening companies' ability to detect and respond to these risks, this bill will help keep Americans protected from the real-world harms posed by compromised AI systems. The Center for AI Safety Action Fund is proud to support this bill," said Varun Krovi, Executive Director, Center for AI Safety Action Fund.

The full text of the legislation is available here.

A one pager on the legislation is available here.

###