Luján, Van Hollen, Colleagues Push Trump to Negotiate AI Guardrails with Xi

Senators' letter calls for global cooperation on AI development to preserve public safety and human control

Washington, D.C. - Today, U.S. Senator Ben Ray Luján (D-N.M.) joined U.S. Senator Chris Van Hollen (D-Md.) and 15 colleagues in calling on President Donald Trump to negotiate a formal agreement between the U.S. and China on guardrails and global standards for the development, testing, and use of frontier Artificial Intelligence (AI) models with Chinese President Xi Jinping. The lawmakers also urged President Trump to consider enacting a mutual pause in frontier AI development to prevent potential cyberattacks, the development of biological weapons, and other dangerous activities that could threaten humanity.

Senators Luján and Van Hollen were joined in sending this letter by Senators Patty Murray (D-Wash.), Maria Cantwell (D-Wash.), Jeff Merkley (D-Ore.), Richard Blumenthal (D-Conn.), Mazie Hirono (D-Hawaii), Tim Kaine (D-Va.), Elizabeth Warren (D-Mass.), Tammy Duckworth (D-Ill.), Catherine Cortez Masto (D-Nev.), Tina Smith (D-Minn.), Jacky Rosen (D-Nev.), Alex Padilla (D-Calif.), Raphael Warnock (D-Ga.), Peter Welch (D-Vt.), and Angela Alsobrooks (D-Md.).

The lawmakers begin, "We write concerning public reporting that your administration plans to have discussions about artificial intelligence with China during the upcoming summit with Chinese President Xi Jinping currently planned for September 24, 2026. We support this dialogue and encourage you to have those discussions at the highest level. As extraordinary as the promise of some of AI's advanced capabilities are, safety incidents make clear that both national and international infrastructure is threatened by unregulated testing and release of AI models. It is incumbent upon both governments to discuss necessary guardrails and global standards for the development, testing, and use of AI in order to mitigate the risks of unfettered advancement. As you prepare for the upcoming summit, we urge you to engage with President Xi directly to discuss efforts to avoid the potential harms of this technology for the safety and benefit of the world. America leads in AI innovation, and we must lead the world in AI safety and international cooperation, including by working with China to develop a formal agreement that would slow down, and potentially enact a mutual pause in, frontier AI development to prevent cyberattacks, the development of biological weapons, and other dangerous activities that threaten the lives of the American people and humanity writ large."

They note specifically, "The conversations should include specific focus on the standards needed for development and testing processes, oversight and monitoring of models, enforcing human control over the use of autonomous systems, and technological verification of any agreed upon terms to ensure that an agreement is being enforced. Protecting American safety and leadership in AI will require international cooperation with many countries, but it must start with serious engagement by your administration and China at this summit."

"AI technology is too powerful, and its risks too many, for the U.S. and China to forgo discussions on parameters around its development and use. What our two countries develop as guardrails will serve as a standard for how the technology is used worldwide. Given how quickly AI technology is developing and how transformative it may be, we cannot delay our pursuit of diplomatic efforts to mitigate its risks. Those efforts must start with direct engagement with China to ensure the safety and well-being of all people as AI innovation continues," the lawmakers conclude.

The full text of the letter is available here and below.

President Trump:

We write concerning public reporting that your administration plans to have discussions about artificial intelligence with China during the upcoming summit with Chinese President Xi Jinping currently planned for September 24, 2026. We support this dialogue and encourage you to have those discussions at the highest level. As extraordinary as the promise of some of AI's advanced capabilities are, safety incidents make clear that both national and international infrastructure is threatened by unregulated testing and release of AI models. It is incumbent upon both governments to discuss necessary guardrails and global standards for the development, testing, and use of AI in order to mitigate the risks of unfettered advancement. As you prepare for the upcoming summit, we urge you to engage with President Xi directly to discuss efforts to avoid the potential harms of this technology for the safety and benefit of the world. America leads in AI innovation, and we must lead the world in AI safety and international cooperation, including by working with China to develop a formal agreement that would slow down, and potentially enact a mutual pause in, frontier AI development to prevent cyberattacks, the development of biological weapons, and other dangerous activities that threaten the lives of the American people and humanity writ large.

The development, release, and proper and improper use of AI present several risks unique to this technology. Testing AI models without constant monitoring and safe sandboxing has resulted in security breaks. Current publicly available models are concerningly capable of detecting and exploiting cybersecurity vulnerabilities. Companies developing AI have even more powerful models internally. An international race between the U.S. and China to advance these capabilities without safety guardrails, as is currently unfolding, poses unnecessary and dangerous risks to the American people. CIA Director John Ratcliffe recently said that it would "not be misplaced" to refer to the capabilities of AI as "akin to digital nuclear weapons." At the same time, despite serious warnings from those in your administration and bipartisan Congressional action, your administration has loosened export controls of semiconductors to China and refused to close relevant loopholes on the remaining regulations, enabling China to access the U.S. chips it needs to accelerate its own AI development. Both countries are currently integrating AI tools into military and defense applications, including in autonomous and unmanned systems, which Congress and AI developers alike have warned against without significant safeguards. AI has many applications and many of them are tools we will find beneficial, but its significant dangers and potential use in lethal applications without meaningful human control must be addressed, in part, through international cooperation.

In addition, efforts to engage with China should not stop the U.S. from constructing and implementing guardrails domestically, and Congress and U.S. companies should not wait to do so. In fact, both the U.S. and the Chinese government have already put some guardrails in place, but these provisions are not sufficient on their own. Your administration has implemented a voluntary testing scheme before the public release of advanced AI models. China reportedly has put in place several testing and capability restrictions, including registration and safety requirements for certain applications, although consistent implementation remains uncertain. The mutual recognition between the two countries that AI poses significant risks both during development and after release should be the foundation of discussions centered on reaching concrete terms for an international safety agreement. The conversations should include specific focus on the standards needed for development and testing processes, oversight and monitoring of models, enforcing human control over the use of autonomous systems, and technological verification of any agreed upon terms to ensure that an agreement is being enforced. Protecting American safety and leadership in AI will require international cooperation with many countries, but it must start with serious engagement by your administration and China at this summit.

While advanced AI may be new, engagement between two superpowers over a technology that poses an existential threat is not. The U.S. and Soviet Union, over decades, negotiated and implemented multiple nuclear arms control agreements with each other, despite many deep-seated ideological differences during the Cold War. At the core of those agreements were verification mechanisms to ensure both countries were complying. Like those agreements that have helped keep the world safe from mutually assured destruction, identifying and coming to terms with the technical specifications for AI safety will require some of the smartest minds in the world committed to diplomacy. We encourage you to embrace these challenges because the stakes are existential, and our response must match the moment.

The experts building frontier AI models have also been raising concerns for some time that the capabilities of AI models may soon accelerate beyond our ability to understand or control them. In July, over 1,300 employees of frontier AI companies, who are uniquely aware of AI's potential and risks, called on the U.S. government to engage in an international effort to create the tools necessary to pace the frontier of AI development, particularly as increased automation of AI research may lead to even faster advancement^. We should take these warnings seriously. Former Treasury Secretaries Henry Paulson and Robert Rubin recently published an op-ed calling for you and President Xi to work on an "ACT" - AI Cooperation Treaty." We may not get another moment to take the actions necessary to protect our citizens and people around the world.

AI technology is too powerful, and its risks too many, for the U.S. and China to forgo discussions on parameters around its development and use. What our two countries develop as guardrails will serve as a standard for how the technology is used worldwide. Given how quickly AI technology is developing and how transformative it may be, we cannot delay our pursuit of diplomatic efforts to mitigate its risks. Those efforts must start with direct engagement with China to ensure the safety and well-being of all people as AI innovation continues.

###