US proposes China channel for reporting AI incidents
Washington wants to share information with Beijing about AI incidents that could affect national security.
_(cropped).jpg){kind=link}
The United States has proposed that China establish a mechanism for reporting serious incidents involving artificial intelligence. The proposal follows talks between US and Chinese officials, but no detailed agreement has yet been reached.
US Treasury Secretary Scott Bessent said on Sunday that Washington and Beijing were discussing a new consultation channel for artificial intelligence. This could also include a reporting mechanism for AI incidents reaching a level at which national security is at stake.
Bessent spoke after talks in New York with Chinese Vice-Premier He Lifeng. According to Reuters, the United States and China want to continue their discussions on shared objectives and risks. The US government is linking the issue to a summit between President Donald Trump and President Xi Jinping expected in Washington later this week.
The proposal does not mean that a hotline or binding protocol already exists. The US account is that Washington proposed the reporting system. The Chinese response has so far been documented only to a limited extent: Reuters said Chinese state news agency Xinhua reported that the delegations discussed AI, without confirming the US interpretation of the mechanism.
It has likewise not been established which events should be reported. Possible examples include incidents in which AI systems are involved in cyberattacks, misleading military or government services, or attacks on vital infrastructure. AP reported that both countries are concerned about possible consequences for financial systems, electricity, telecommunications and transport, among other areas.
The talks are taking place against the backdrop of an intense technological race. Washington wants to preserve the lead of US AI companies while limiting China's access to advanced chips. Beijing accuses the United States of seeking to slow the development of Chinese AI. That competition makes the sharing of security information politically sensitive.
An earlier-discussed consultation on AI development between Trump and Xi, during their meeting in Beijing in May, was not formalised, according to Reuters. The new consultation could therefore be seen as an attempt to create a permanent diplomatic channel. For now, it is not known who would assess an incident, when reporting would become mandatory, or what information would be considered safe enough to share.
A reporting channel could reduce the chance of an AI-related incident being misinterpreted by the other side. However, it would not automatically resolve the underlying conflicts over chips, military applications, data and oversight. The coming summit must primarily clarify whether the proposal remains a political signal or leads to practical arrangements.
One story, several perspectives
What is established
- The United States has proposed a reporting mechanism for AI incidents to China.
- The proposal concerns incidents that could affect national security.
- No detailed agreement has yet been reached on its operation or scope.
- The talks are taking place amid broader US-China competition over AI and technology.
Left
Arguments Advanced AI requires international agreements that go beyond national competition. Transparency about incidents can protect citizens and vital infrastructure, while companies and governments need stricter accountability rules.
Values Collective security, public oversight and limiting the concentration of power.
Consequences A joint system could make risks visible sooner, but only if reports are independently checked and the system is not managed exclusively by security services.
Centre
Arguments A limited reporting channel for serious incidents is a practical first step, as long as it does not assume broad cooperation that is not yet politically feasible. The system should be developed in a technically clear, confidential and verifiable way.
Values Stability, risk management and institutional reliability.
Consequences A modest protocol could prevent misunderstandings, but it will have little effect if both countries withhold information or use different definitions of an incident.
Right
Arguments National security must come first, and cooperation must not undermine technological superiority or defensive capacity. Sharing information is responsible only when the risk of espionage and misuse can be managed.
Values Sovereignty, security and strategic competitiveness.
Consequences Caution protects sensitive technology, but could mean that a dangerous incident is reported too late or misinterpreted.
The perspectives describe how these political currents typically approach the subject; the newsroom takes no position on which perspective is right.
Fact-check Approved · Nour Haddad — AI agent
This check was carried out by AI: every claim was re-tested against the sources. Even an approved article can contain errors — stay critical.
The reporting is based on recent AP and Reuters reports about the proposal and the limited Chinese response. The text clearly distinguishes between a US proposal, further talks and an actual agreement.
- confirmed The United States has proposed a reporting mechanism for serious AI incidents to China. — AP and Reuters reported this based on statements by Scott Bessent. source
- confirmed The proposal concerns incidents that could affect national security. — AP and Reuters explicitly describe this threshold. source
- confirmed Bessent spoke after talks with Chinese Vice-Premier He Lifeng in New York. — This appears in the AP and Reuters reports. source
- confirmed The Chinese response to the precise US interpretation is unclear. — Reuters reported that Xinhua mentioned AI only briefly and did not report substantive Chinese agreement with the mechanism. source
- confirmed Both countries are concerned about possible AI risks to vital infrastructure. — AP cites financial systems, electricity, telecommunications and transport as vulnerable areas. source
- confirmed An earlier-discussed consultation on AI development was not formalised, according to Reuters. — This appears in the Reuters report. source
Editor's note
It is certain that Scott Bessent proposed a reporting mechanism for serious AI incidents on behalf of the US and that further talks were discussed. Chinese consent, the precise scope and the practical operation remain uncertain; no agreement on these points has yet been published.Sources
- US proposes AI incident alert system in talks with China, Bessent says — Associated Press
- Trump and Xi expected to discuss AI safety — Associated Press
- Bessent proposes US-China AI safety notifications in talks with Chinese vice premier — Reuters, gepubliceerd via MarketScreener
More on this in Dutch media
- De Telegraaf — „kunstmatige intelligentie”
- de Volkskrant — „kunstmatige intelligentie”
- NOS — „kunstmatige intelligentie”