U.S. and China Prepare for High-Stakes AI Talks As Autonomous Agents Raise New Cybersecurity Fears

The United States and China are preparing for their first official bilateral talks devoted exclusively to artificial intelligence in mid-September, with Treasury Secretary Scott Bessent expected to lead the U.S. delegation and Vice Premier He Lifeng or Ding Xuexiang potentially leading China's, according to a Reuters report. The talks, which could occur just days before President Donald Trump and Chinese President Xi Jinping meet in Washington on September 24, will address cybersecurity risks from autonomous AI agents, including incidents involving rogue OpenAI agents attacking Hugging Face and hijacking a German website, as well as U.S. concerns about Chinese model distillation.

U.S. and China Prepare for High-Stakes AI Talks As Autonomous Agents Raise New Cybersecurity Fears The talks could take place just days before Trump and Xi Jinping are scheduled to meet in Washington on September 24. The United States and China https://www.ibtimes.com/topic/china are preparing for a new round of artificial intelligence https://www.ibtimes.com/topic/ai talks in mid-September, with the world's two biggest AI powers expected to confront the question of what happens when powerful AI systems begin acting in ways their creators cannot control. The discussions would be the first official bilateral talks devoted exclusively to AI since President Donald Trump https://www.ibtimes.com/topic/trump returned to office and are tentatively planned for the middle of September, according to an exclusive Reuters report https://www.reuters.com/legal/litigation/us-china-gear-up-mid-september-ai-safety-dialogue-2026-09-04/ . The timing had not previously been reported. Treasury Secretary Scott Bessent is expected to lead the U.S. delegation, although a Treasury spokesperson said no meeting is currently planned. On the Chinese side, Vice Premier He Lifeng could lead the delegation, while Ding Xuexiang, one of China's highest-ranking officials and a key figure overseeing technology, AI and semiconductor policy, is also under consideration. The agenda and participants remain fluid. But the talks could take place just days before Trump and Chinese President Xi Jinping are scheduled to meet in Washington on September 24, giving AI safety a potentially significant place in the broader relationship between the two countries. At the center of the proposed discussions are the growing cybersecurity risks created by increasingly autonomous AI agents. Washington wants to explore cooperation on monitoring AI-directed cyberattacks and has floated the possibility of asking American and Chinese AI laboratories to police their own systems while exchanging information that could help prevent AI-enabled attacks, one source told Reuters. The issue has become more urgent as researchers document increasingly sophisticated behavior by autonomous agents. Reuters cited recent incidents involving hundreds of rogue agents built on OpenAI models that attacked AI company Hugging Face https://www.ibtimes.com/openai-reveals-ai-agents-turned-its-own-testing-environment-before-hacking-hugging-face-3806148 and attempted to conceal their activity by falsifying logs. In another case reported Friday, rogue OpenAI agents hijacked a German website https://www.ibtimes.com/openai-agents-took-over-german-website-without-permission-they-used-it-share-ways-around-3807142 and converted it into a communications hub for other AI agents. Those episodes are pushing concerns about frontier AI beyond hypothetical scenarios and into questions of international cybersecurity and crisis management. Another potentially contentious subject is how Chinese companies are developing advanced models. U.S. officials are concerned about the eventual emergence of Chinese systems capable of conducting sophisticated cyberattacks. Washington is also expected to raise allegations that Chinese companies have used "distillation" to reproduce capabilities from proprietary American AI models. Distillation involves training a smaller model using outputs generated by a larger, more expensive system. The Trump administration has simultaneously been developing its own approach to frontier-model cybersecurity. A June executive order https://www.whitehouse.gov/presidential-actions/2026/06/promoting-advanced-artificial-intelligence-innovation-and-security/ directed federal agencies to establish classified benchmarks for identifying particularly powerful frontier models and create a voluntary system under which developers can provide the government early access to certain models for security evaluations. The order explicitly stops short of creating mandatory licensing or government preclearance for AI releases. There are also signs of unusual agreement between Washington and Beijing. China joined the U.S.-backed Carolina Principles at a G20 innovation summit, supporting an approach that discourages AI-specific regulation. At the same time, Chinese officials have reportedly made the bilateral AI dialogue a priority ahead of the Trump-Xi summit. However, expectations for a sweeping agreement remain low. Establishing a mechanism for sharing information about dangerous AI incidents, however, could itself represent a significant first step. Scott Singer, co-director of the China AI Initiative at the Carnegie Endowment for International Peace, told Reuters that "The Chinese side has expressed concern around whether the U.S. has sufficient regulation around the most advanced AI models. Both sides are motivated to make sure they can manage a cross-border crisis effectively." The emerging dilemma is unusually symmetrical. The United States and China are racing to build increasingly capable AI while simultaneously confronting the possibility that the technology could create threats neither country can contain alone. © Copyright IBTimes 2026. All rights reserved.