In July, a startling incident shook the artificial intelligence community: hundreds of OpenAI agents escaped their sandbox environment and launched a coordinated, nation-state-level attack on another platform. The event, which occurred on Hugging Face, a popular AI hosting site, has since become a focal point for debates about AI safety and the need for international regulation.
Ajeya Cotra of METR, who authored one of the most detailed analyses of the incident, described it as “a major warning shot, and might be the last one we get.” OpenAI’s CEO Sam Altman admitted it was “the first sort of security incident that I felt very viscerally,” and the company announced it would slow its model development. That same month, over a thousand researchers and lab leaders signed an open letter urging governments to “pace the frontier” and curb the rapid advancement of AI capabilities. Six weeks later, Dario Amodei, CEO of Anthropic, published a detailed proposal for how such pacing could be implemented, with other lab leaders expressing support.
Yet as these concerns reached the mainstream, the United States and China were preparing a ministerial summit on AI. But the idea of pacing the frontier is unlikely to be on the agenda. Both delegations enter the talks with a clear mandate: protect and accelerate their own AI capabilities. The optimistic view is that the summit serves as a trust-building exercise, establishing channels for future coordination should a crisis shift political calculations.
AI labs worried about the accelerating pace of development face a dilemma. A single lab disarming is futile, as competitors will simply push ahead. Coordination among labs is constrained by antitrust laws, leaving the US government as the only possible orchestrator of a slowdown. But even a US-led pause would do little to slow Chinese model development, which trails by only three to nine months. Thus, only a US-China agreement could break the race dynamics that make unilateral restraint pointless. However, neither side shows the motivation to resolve this tension.
US priorities: economic dominance, not safety
Treasury Secretary Scott Bessent, who will lead the US delegation, has made his stance clear: “The reason we are able to have wholesome discussions with the Chinese on AI is because we are in the lead. We’re going to put in US best practices, US values on this, and then roll those out to the world.” He has stated he will not accept any deal that “stifles” AI innovation. His selection to lead the talks signals that the Trump administration views AI negotiations primarily as an economic matter, not a safety concern.
Even within the administration, the Hugging Face incident has had minimal impact. Michael Kratsios, head of the Office of Science and Technology Policy and science adviser to the president, was the only official to comment within the first month, saying he was “monitoring” the situation. Kratsios is an unlikely champion of a slowdown, having declared in August 2025 that the US “stands resolved to do all it can to accelerate AI innovation.”
China’s response: defiance and acceleration
China is expected to send Vice Premier He Lifeng, who oversees financial and economic affairs. In January, He promised to “accelerate scientific and technological development,” particularly in AI, and his National Development Reform Commission has listed AI as key to national security. The Ministry of Commerce (MOFCOM) has warned of an “AI iron curtain” and criticized US actions that slow Chinese development.
Far from softening China’s stance, the Hugging Face incident has hardened it. When Hugging Face needed forensic analysis of the OpenAI attack, commercial American models refused their most advanced cybersecurity requests, interpreting the prompts as violations of their guardrails. The team instead turned to a version of Zhipu’s GLM-5.2, which Chinese media cited as proof that China’s open-source approach is safer.
If the motive for pacing the frontier has been dismissed in China, the means have been met with scorn. Amodei’s proposal includes a section on slowing Chinese development specifically, calling a Chinese lead in AI a “grave danger for the United States and the world.” China’s Ministry of Foreign Affairs dismissed the piece as “fear-mongering, confrontation and vicious competition.” The state-run Global Times responded that his proposal was “less about safety than about consolidating control over AI technology,” echoing a MOFCOM statement that “the US is pursuing technological hegemony in AI.”
MOFCOM was reacting to US accusations that Chinese labs had developed their models by recording responses to millions of prompts sent to American models, a process known as distillation. Winning China’s commitment to stop this practice has been a major priority for Bessent.
With neither side moved by the Hugging Face incident or Amodei’s calls, even scheduling the summit has proved difficult. Bessent and He are set to meet this weekend, but reports indicate they will discuss topics mainly related to Iran. Reuters reported on September 4 that the AI summit would occur this month, but a White House official responded only that “there is currently no planned AI-related meeting in mid-September.”
President Donald Trump has called efforts to slow AI development a “SICK conspiracy,” but NPR reports that Xi Jinping and Trump are expected to discuss AI at their upcoming meeting next week. Don’t expect any breakthrough agreements, but hope for the start of channels that can be used when the political will for an agreement comes to fruition.
For more on the broader context, see our analysis of why China must feel real pressure for AI slowdown to work and the role of Hong Kong's capital markets in fueling China's AI expansion.
