The United States and China are preparing to hold an AI safety dialogue in mid-September, according to people familiar with the matter, as the two governments seek common ground on the rapidly evolving risks posed by artificial intelligence. The talks come amid escalating concern in Washington and allied capitals that advanced AI systems could be repurposed to launch sophisticated cyberattacks against critical infrastructure, financial networks, and government databases.
One of the central items on the proposed agenda, according to a source who spoke on condition of anonymity, is cooperation on monitoring cyberattacks that are directed or enhanced by AI. The US side has also floated a proposal under which leading American and Chinese AI laboratories would effectively "police themselves," voluntarily exchanging information about emerging threats and the capabilities of frontier models. The aim, the source said, is to build an early-warning channel before malicious use of the technology escalates into a cross-border crisis.
ALSO READ | Apple’s First Foldable iPhone Duo Set for Tonight’s Launch with Leaked Color Lineup and Pricing
The dialogue reflects a broader, if cautious, effort by the Biden-era framework for US-China technology engagement to continue functioning despite deep trade frictions and ongoing export-control disputes over advanced semiconductors. Both governments have publicly acknowledged that AI safety is one of the few areas where bilateral talks remain viable, even as broader strategic competition has intensified. A formal announcement of the date, venue, and level of representation is expected in the coming days.
The proposal to involve private AI labs directly in self-policing arrangements would mark a notable departure from purely government-to-government engagement and signals Washington's preference for industry-led guardrails supplemented by state diplomacy. If accepted, the framework could see major American labs including OpenAI, Anthropic, Google DeepMind counterparts, and Chinese players such as Baidu, Alibaba, and ByteDance sharing threat intelligence on AI-enabled hacking tools, model jailbreaks, and dual-use research.
Analysts caution, however, that previous US-China dialogues on emerging technologies have struggled to translate commitments into verifiable action, particularly on issues tied to national security. Skeptics note that Beijing has historically resisted transparency mechanisms that could expose the operational footprint of state-linked cyber units. The success of the September talks may therefore hinge on whether the self-policing arrangement can be structured in a way that addresses Chinese concerns about sovereignty while delivering credible risk reduction for Washington.
