US and China keep talking about AI safety despite rivalry

In short: The US and China are competing on AI, but they are also keeping narrow channels open to reduce shared risks, especially around military use.

What's going on

US and Chinese leaders have continued to pursue limited cooperation on artificial intelligence, even while both countries try to outpace the other. The idea is simple, some AI risks affect everyone, so neither side can fully handle them alone.

One clear example is government talks. In November 2023, President Joe Biden and President Xi Jinping set up a government to government channel focused on AI risks and safety. In 2024, this led to a notable commitment, both sides agreed to prevent AI from controlling nuclear weapons systems (meaning software should not be allowed to make launch decisions).

Cooperation is also showing up in research. Joint US China AI research has grown sharply over time, reaching 9,660 shared papers in 2021, which is more than any other country pair. Workshops have continued too, including a May 2025 event at Peking University where people discussed ideas for global rules and oversight.

Both countries are also engaging in international AI safety discussions involving leading scientists. China has created an organization meant to match US AI safety institutes, with a focus on making sure advanced AI systems do what people intend (like a safety check before letting a machine handle important tasks).

What to watch

The biggest question is how long these safety efforts can continue while the broader relationship stays tense. The US has tightened restrictions on advanced AI chips and added some groups to blacklists, while China is pushing its own national AI plans. Watch for whether future summits and military focused talks keep moving forward, especially under a US approach that may favor domestic control and faster innovation over shared global standards.

Source: NYTimes