Trump’s September 29 remarks reject deeper AI-risk cooperation
President Donald Trump used September 29 remarks to reject deeper work with China on artificial-intelligence risks, framing such cooperation as a potential threat to the U.S. position. South China Morning Post reported on September 30 that Trump said he did not want to work with China on addressing AI risks. He argued that the United States was leading China and other countries and that preserving that lead required maintaining the current course. His comments presented a stated view of competitive advantage; the reports do not supply a quantified comparison of national AI capabilities.
The remarks place a clear limit on how far Trump appears willing to take U.S.-China engagement on technology risks. The Associated Press separately reported on September 29 that he did not want the United States to cooperate too much with China. Together, the accounts show Trump linking restraint on cooperation to his assertion of U.S. leadership, while leaving the previously announced bilateral mechanisms unaddressed. Neither account describes the September 29 comments as a new treaty, a formal policy document, or a move to cancel those mechanisms.
SCMP also reported that calls for U.S.-China collaboration had grown in recent weeks around concerns that frontier AI could be misused for cyberattacks and bioweapon development, in the context of incidents involving OpenAI and Anthropic. That is SCMP’s reported context for the debate, rather than evidence that Washington and Beijing had embraced a broader joint safety framework.
Trump’s language also centered on superintelligence as a winner-take-all contest. SCMP used that term in reporting his remarks, while the official bilateral terminology differs. The difference is important because the White House gives its formal arrangement a distinct name and scope. Treating Trump’s competitive rhetoric as a measure of how he views the AI race does not establish a shared U.S.-China policy on risk management.
The White House describes a limited incident communication channel
The White House fact sheet dated September 25 says the United States and China established the U.S.-China Super Intelligence, or SI, Dialogue. The fact sheet says the two leaders adopted the term super intelligence in place of artificial intelligence for applicable emerging technologies and that the dialogue will exchange views on their risks and benefits. It says the countries committed to creating an SI incident contact channel, with the next exchange due by November 2026.
AP described the same arrangement as a channel for handling AI-related incidents, agreed during Xi Jinping’s Washington visit in the prior week. The official fact sheet does not describe a joint research program, shared safety rules, or binding bilateral requirements. Its stated elements are an exchange of views and an incident communication channel. The cited accounts do not report that the dialogue or channel was withdrawn after Trump’s September 29 remarks.
The White House fact sheet sets an exchange by November 2026 as part of the SI Dialogue’s stated timetable.
The labels in the available accounts should not be collapsed into a single claim about AI safety. SCMP’s report concerns work on AI risks and uses AI-safety framing in its headline. AP describes AI-related incidents. The White House describes risks and benefits related to SI. Those formulations cover overlapping concerns, but the available accounts do not describe one broad, jointly agreed regulatory program.
EastFrontier previously covered the September U.S.-China AI dialogue and the summit’s AI-dialogue outcome. Those archived accounts are background context; the present article relies on the cited September 25 White House fact sheet and September 29–30 reporting.
A six-company U.S. accord remains separate and voluntary
Trump also announced a domestic accord involving leaders of six major U.S. technology companies, a development distinct from the bilateral channel. SCMP characterized it as voluntary. Reuters reported that Trump released the one-page document on Truth Social and described it as morally binding. Reuters identified Trump and the heads of Google, Anthropic, Meta, OpenAI, X, and NVIDIA as its signers. The signatory list does not include Amazon or Microsoft.
Reuters said the accord calls for external assessment of companies’ AI-control and monitoring systems, as well as an independent board committee to oversee and receive audit and control reports. It also lays out internal controls for capabilities and alignment during training and deployment, including controls related to cybersecurity, biosecurity, and chemical threats. The document further calls for an internal team to make sure the controls, monitoring, and detection systems operate as intended. These are requests set out in a company accord, not current federal requirements.
The document’s own forward-looking language leaves the possibility of later legal or regulatory action open. Reuters reported that it says participating companies will meet regularly to develop standards and best practices and that codifying the steps into laws or regulations could make sense over time. The six-company arrangement remains voluntary. The document identifies processes it asks participants to adopt and says future legal or regulatory steps could be considered.
AP’s account of the White House lunch lists Anthropic Chief Executive Officer Dario Amodei, OpenAI President Greg Brockman, Amazon founder Jeff Bezos, Elon Musk, NVIDIA Chief Executive Officer Jensen Huang, Google Chief Executive Officer Sundar Pichai, Meta Chief Executive Officer Mark Zuckerberg, and Microsoft Chief Executive Officer Satya Nadella among the attendees. Reuters’ signer list is narrower. The accounts therefore distinguish a voluntary domestic company accord from the White House’s separately described bilateral SI Dialogue.