# Trump's AI race obsession may backfire with China > US-China talks on AI safety risk becoming theater if they ignore real risks and focus on winning a nonexistent race. - Title: Trump's AI race obsession may backfire with China | Shortify - Summary: US-China talks on AI safety risk becoming theater if they ignore real risks and focus on winning a nonexistent race. Framing AI as a race to win pushes China… - Keywords: trump, xi jinping, us-china relations, ai safety, ai governance, export controls, scott bessent, sam altman, artificial intelligence, geopolitics, tech competition, notification system - Source: Ars Technica — https://arstechnica.com/tech-policy/2026/09/china-silent-as-us-touts-plan-for-ai-safety-alerts-that-omits-tech-experts - Published: 2026-09-23T20:52:28+00:00 - Read time: 3 min - Topics: trump, xi jinping, us-china relations, ai safety, ai governance, export controls, scott bessent, sam altman, artificial intelligence, geopolitics, tech competition, notification system ## US and China finally talk AI safety Trump and Xi are meeting this week with AI at the top of the agenda. Treasury Secretary Scott Bessent announced both sides discussed creating a notification system for AI safety alerts. ## Alert system would create direct line The proposed mechanism would let the US or China send alerts when the other's AI systems pose threats or behave unpredictably. It could also allow both sides to share their vision of common goals. ## Trump keeps tightening China's tech access Trump has steadily increased export controls blocking China from AI technology. Congress is also mulling bills to cut China off from more chips and hardware, raising tensions before talks. ## Trump accuses China of AI theft Two weeks before talks, Trump's Justice Department accused six of China's leading AI makers of stealing from US frontier labs. China has strongly disputed the allegations. ## China stays quiet on alert system China has not publicly acknowledged the proposed notification system. Even Trump's Truth Social posts declaring America must dominate China on AI didn't keep China from showing up to talks. > WHOEVER WINS AI, WINS! We are leading China, and all others, and will continue to do so ## Plan lacks technical expertise The alert system involves no officials with technical expertise to assess the urgency of risks. People familiar with the plan told Politico it appears superficial. > the scientific, technical term is BS ## Experts see red flag in design Sam Bresnick from Georgetown's Center for Security and Emerging Technology says it's a red flag that no tech experts monitor the channel, especially for military risks. > It's better than nothing for sure, but I don't have a lot of confidence that that is going to make a major difference going forward ## Much more cooperation needed Su Lian Jye from Omdia says the alerts may promote transparency but won't reduce wariness. Both countries must agree on risk classifications, notification triggers, and response protocols. ## Khanna wants bans on dangerous AI US Representative Ro Khanna, ranking Democrat on the House Select Committee on US-China competition, says Trump should push Xi for an international agreement banning recursive self-improving AI and safeguards against bioweapons. ## China could share audit expertise Alvin Wang Graylin from Stanford says China regularly audits AI models and could share knowledge from hundreds of audits, while the US has only just begun. Both countries have much to learn from each other. ## China doesn't fear AI the same way In China, there's not the same public fear about uncontrollable AI because the state feels it can control AI. This may change as China's advanced AI catches up to the US. ## China sees safety talk as containment When the US ties AI safety discussions to export controls and keeping America ahead, China views it as suppressing Chinese AI rather than genuine risk management. This suspicion could render the alert system useless. > I think there are reasons [for China] to view this notification mechanism with suspicion ## AI race narrative is deeply flawed Graylin warns the US is caught up in a deeply flawed narrative about winning an AI race. In reality, there are two races: one for intelligence and one for market dominance. China is winning on adoption of open models. ## No permanent advantage possible US-China capability gaps at the frontier last only months with no competitive moat. Even if the US achieves artificial superintelligence first, the advantage could evaporate in weeks or months without deployment. > US-China capability gaps at the frontiers of AI technology typically last only a few months ## Real threats are smaller models The most dangerous AI models are often smaller specialized ones that non-state actors could use for biological, chemical, or cyber weapons. The US-China conflict narrative prevents the cooperation needed to address this real threat. > The most dangerous models are often those smaller specialized ones that could be created or utilized by non-state rogue actors ## Trump should ease export controls Graylin suggests Trump could loosen export controls to only block genuine military and chemical-biological-radiological-nuclear threats, not all advanced technology. This might let China deploy safer AI products. ## There is no way to compete to safety Graylin says the US cannot manage the risks of the most powerful technology in human history while refusing to talk to China. There is no way to compete yourself to a safer world. > There is no way to compete yourself to a safer world ## Key takeaway Framing AI as a race to win pushes China away from cooperation on genuine safety threats like bioweapons and cyberattacks.