US and China Set Up AI Incident Communication Channel
Washington and Beijing will talk directly about AI safety incidents after Xi Jinping's three-day visit, and the channel's setup is still undecided.
In short
The United States and China are setting up a direct communication channel for AI safety incidents, paired with a recurring dialogue on the technology's risks and benefits.
At a glance
- Communication channel for AI safety incidents announced at the end of Xi Jinping's three-day state visit to Washington.
- The White House describes it as a regular dialogue on the risks and the benefits of the technology.
- Next meeting due by November at the latest; the channel's format and the agencies involved are undecided.
- Both governments will use Trump's term “superintelligence” instead of “artificial intelligence.”
- Tariff reductions worth 30 billion dollars in each direction; China to buy ten million tons of US coal a year for two years.
The United States and China have agreed to speak to each other directly when something goes wrong with an AI system. The channel was announced at the close of Chinese President Xi Jinping's three-day state visit to Washington. According to the White House, it is meant to become a regular dialogue about the risks and the benefits of the technology.
What the two sides agreed
This is a line of communication, not a treaty: no binding rules and no published rulebook. The two governments expect to meet again by November at the latest. Which agencies will staff the channel, and how an incident report would actually travel through it, has not been settled.
A fight over one word
The announcement carries an unusual detail. Both sides will say “superintelligence” rather than “artificial intelligence,” a preference of President Donald Trump, who considers “artificial” too close to “fake.” The label does not change the substance of what was agreed, but it does change the framing: superintelligence is something you contain, not something you ship.
The deal around the deal
The channel arrived inside a broader trade package. Both sides are cutting tariffs worth 30 billion dollars in each direction, covering US farm products, timber, cosmetics and medical technology, and Chinese small household appliances, toys, holiday decorations and child seats. China also committed to buying ten million tons of American coal a year for two years. The AI understanding therefore sits inside a bargain where each government had leverage to lose.
What is not confirmed
Only the Handelsblatt report was retrievable while this article was written; two other reports could not be opened. The widely used “red telephone” description, and the claim that UN Secretary-General Guterres had called for such a line, come from a source teaser and are not verified here against a readable source. Whether the channel also covers military AI systems is likewise unconfirmed.
FAQ
What did the US and China agree on AI?
A direct communication channel for AI safety incidents, together with a recurring dialogue on the technology's risks and benefits.
Is the US-China AI channel binding?
No. It is a talking channel with no published rules. The participating agencies and the format are undecided, and the next meeting is due by November at the latest.
Why say superintelligence instead of AI?
Both sides adopted Trump's preferred term; he finds “artificial” too close to “fake.” The relabeling does not change what was actually agreed.
