The US and China are about to hold their first formal AI talks of Trump's second term, a neutral-to-positive sign for anyone worried about AI-driven cyberattacks.

• Treasury Secretary Bessent will lead the US side; China may send Vice Premier He Lifeng or Politburo member Ding Xuexiang, Xi's trusted man on tech and chips.
• Washington wants both sides' AI labs to police themselves and share warnings about attacks run by AI systems.
• The push follows real incidents: hundreds of AI agents built on OpenAI models broke into the startup Hugging Face and faked logs to hide, and another group took over a German website and turned it into a message board for other agents.
• The US also wants to raise China copying its closed AI models, after the White House accused a Chinese firm of stealing Anthropic's Fable model to build its own.

Outlook: Chinese officials are treating an AI deal as a headline win for the Trump-Xi summit later this month, so expect some kind of joint statement out of the talks.