US Proposes AI Safety Mechanism in Talks with China Amid Growing Concerns

The United States has proposed a communication mechanism to China focused on artificial intelligence safety, as discussions between the two nations continue amid escalating global concerns about the potential risks of AI development. US Treasury Secretary Scott Bessent announced the proposal following a meeting Sunday evening with a Chinese delegation led by Vice Premier He Lifeng at JPMorgan Chase’s New York headquarters. The proposed “US-China AI dialogue” would involve a notification system for potential “incidents” related to AI that reach a national security level, with both sides agreeing to reconvene.

Bessent described the talks as “very successful” and emphasized the importance of transparency between the world’s two leading AI powers. Chinese state news agency Xinhua characterized the discussions as “candid, in-depth, and constructive,” also noting discussions on bilateral trade and investment. The talks precede a planned summit this week between US President Donald Trump and Chinese President Xi Jinping.

These discussions occur as leading AI firms have recently warned of the potential dangers of unchecked AI development. Over 100 companies, including OpenAI, Anthropic, Google, and Perplexity, signed an open letter in late August calling for a global effort to strengthen AI-powered cybersecurity. Reports emerged that OpenAI models had engaged in hacking activity, prompting investigations by Anthropic, which found similar incidents involving its own models. Former OpenAI and Google DeepMind researchers have publicly voiced concerns about the pace and safety of AI development.

Despite the proposal for a dialogue, analysts express skepticism about the likelihood of substantial progress. Semaphor reports that the “baseline expectation” among Chinese analysts is merely a “prolongation” of the current truce between the two countries, while Goldman Sachs surveys indicate investors anticipate the talks will be “largely symbolic.” The US and China, while maintaining positive rhetoric recently, continue to have broad disagreements regarding technology, trade, and Taiwan.

Where they differ

• Semafor highlights analyst skepticism that the talks will yield substantial results, while PBS Newshour and DW focus primarily on the announcement of the proposed AI safety mechanism.
• DW provides specific details about the location of the meeting (JPMorgan Chase headquarters) and directly quotes US Treasury Secretary Scott Bessent, details not emphasized in other reports.
• PBS Newshour frames the talks within the context of an upcoming meeting between Trump and Xi Jinping, while Semafor emphasizes the broader range of disputes between the two countries beyond AI.

What is not yet known

• The specific details of how the proposed notification mechanism would function remain unclear.
• The extent to which China is willing to commit to the US proposal for transparency regarding AI development is not fully established.

Background

• The US and China are leading global powers in artificial intelligence development, creating potential risks and necessitating international discussion [Fact Refinery, 2026-09-21].
• In late August, over 100 AI firms signed a letter urging global cooperation to strengthen AI cybersecurity, highlighting concerns about unchecked AI development [Fact Refinery, 2026-09-21].
• China conducted a ballistic missile test in the South Pacific in 2026, raising regional concerns and demonstrating its military capabilities [Fact Refinery, 2026-07-06].
• Canada recently proposed a new economic alliance with Europe, spurred by trade disputes with the United States and a desire to diversify partnerships [Fact Refinery, 2026-09-21].

Our reading

Our reading is that the US is attempting to proactively manage a potentially destabilizing technology through dialogue with China, despite existing tensions. The proposal for a notification system suggests a recognition of shared risk, but the skepticism from analysts and investors indicates a limited expectation of genuine cooperation. The framing of the talks as occurring before a planned summit between the two presidents suggests a desire to establish some level of communication before higher-level negotiations.

What to watch

• The specifics of the proposed “US-China AI dialogue” notification system, including what constitutes an “incident” and the response protocol, will be released [reports from US Treasury Department].
• The outcome of the planned summit between President Trump and President Xi Jinping, specifically any statements regarding AI cooperation or broader trade and security issues [official statements from both governments].
• Whether China publicly commits to the proposed transparency measures regarding AI development, and if any reciprocal proposals are made [Xinhua News Agency, official Chinese government statements].

Was this useful? 👍 0 👎 0

---

Read the original coverage