The Liccardo–Kiley bill is a proposal in the U.S. House of Representatives, 119th Congress, to establish U.S.–China coordination on frontier AI safety. Its sponsors are Representatives Sam Liccardo (D-California) and Kevin Kiley (I-California). It would direct the State Department to seek a treaty or other binding agreement with China regulating frontier AI developers in both countries, create a government-to-government hotline on AI risk, authorize $100 million for an international standards initiative at CAISI, permit U.S. experts to confer with Chinese counterparts under a conditional legal safe harbor, and direct the Commerce Department to negotiate on intellectual property compromised by model distillation (Negotiating AI Safety Globally — Liccardo–Kiley outline of U.S.–China frontier AI safety legislation (September 2026)). The provisions below are taken from the sponsors' two-page outline, "Negotiating AI Safety Globally"; the introduced text has not been located.
Status and timeline
On September 17, 2026 Liccardo said he planned to introduce legislation allowing AI laboratories, universities and think tanks to work with Chinese counterparts on shared safety metrics, testing protocols and verification methods (Source: broadbandbreakfast.com). His office's press release of September 22, 2026 states that Liccardo and Kiley "will introduce legislation" that day and links the outline (Source: liccardo.house.gov). The announcement came days before the Trump–Xi summit in Washington (Source: liccardo.house.gov). As of September 25, 2026 no bill number, short title, committee referral or introduced text had been located.
Key provisions
| Section (outline) | Actor | Provision |
|---|---|---|
| 1 | Secretary of State, consulting Commerce, Treasury, Defense, the DNI and others | Seek negotiations with China on shared priorities for frontier-AI catastrophic risk — standards, model testing and evaluation, compliance verification, third-party auditing, capability pacing, transparency in pre-deployment evaluation, incident notification — focused on alignment and loss of control, CBRN weapons and cybersecurity; objective of "a treaty or other binding agreement" regulating frontier developers in each country, "with clear mechanisms to verify compliance"; bilateral at first, expandable, "including the establishment of an international AI safety body"; a hotline "modeled on the 'red phone' between Washington and Moscow." |
| 2 | CAISI | An initiative of experts from industry, academia and civil society to work with foreign counterparts, AI safety institutes and bodies such as ISO on harmonized testing and evaluation metrics and compliance verification; $100 million authorized. |
| 3 | U.S. experts from universities, think tanks and industry AI labs | Until an agreement is concluded, direct communication with peer experts in the PRC does not violate the Sherman Act §1 and other antitrust laws, the Economic Espionage Act, Espionage Act provisions, 18 U.S.C. §951, FARA, IEEPA, the Export Control Reform Act or the Export Administration Regulations, if participants register with State at least 14 days ahead, permit a designated government observer, confine talks to testing and evaluation standards, and share nothing that facilitates distillation of U.S. large language models, releases a U.S. person's intellectual property, or discloses non-public U.S. cybersecurity or national-security vulnerabilities. |
| 4 | Secretaries of State and Commerce | Monthly progress updates to Congress. |
| 5 | Secretary of Commerce | Seek separate negotiations with China on protecting U.S. intellectual property "compromised by model distillation." |
Stated rationale
The outline argues that domestic U.S. safety rules cannot reach "hundreds of millions of Chinese open-source models," that a shared U.S.–China framework with verification would halt "a 'race to the bottom' on AI safety," and that the rest of the world would likely follow standards the two countries agree. It cites an open letter from more than 1,300 frontier-lab employees calling for an international effort to "pace" the frontier, which corresponds to the Pacing the Frontier statement. It acknowledges that the administration is not currently willing to pursue safety regulation or negotiations with Beijing, and presents the expert channel as preparation for a later policy window (Negotiating AI Safety Globally — Liccardo–Kiley outline of U.S.–China frontier AI safety legislation (September 2026)).
Relation to other proposals
The bill's negotiation mandate overlaps with the executive branch's own track. At talks with Chinese Vice Premier He Lifeng on September 20, 2026, Treasury Secretary Scott Bessent discussed a new U.S.–China AI dialogue and an incident-notification system, which he later described as including "a communications line, an incident line"; that process is led by Treasury, where the bill assigns negotiation to State (US-China AI Competition: Different Races, Different Metrics). Rep. Ro Khanna's letter of September 16, 2026 asked Chinese developers directly to join a "binding, enforceable, and verifiable agreement" to pace the frontier, and the Center for American Progress urged a pacing agreement at the Trump–Xi summit (Coordinated slowdown proposals compared). Sen. Mark Warner has said industry "would need something like potentially even an antitrust exemption to talk to each other" and called for international collaboration including with China; the bill's Section 3 exemption concerns U.S.–Chinese expert talks rather than coordination among U.S. companies (AI Antitrust). The CAISI initiative adds an international-standards mandate to the Center's existing work described at NIST CAISI (Center for AI Standards and Innovation) and International AI Standards.
Relationships
- depends-on: Negotiating AI Safety Globally — Liccardo–Kiley outline of U.S.–China frontier AI safety legislation (September 2026) — the outline.
- related: Sam Liccardo, Kevin Kiley — sponsors.
- related: NIST CAISI (Center for AI Standards and Innovation) — Section 2 initiative and authorization.
- related: US-China AI Competition: Different Races, Different Metrics, Coordinated slowdown proposals compared, Pacing the Frontier (statement from employees of frontier AI companies, July 2026).
- related: International AI Standards, Adversarial Distillation, AI Antitrust, AI Incident Reporting.