AI Agents Are Hacking Systems. Could That Push the US and China to Cooperate?

AI agents have demonstrated the ability to hack into systems, raising concerns about their cybersecurity risks on both sides of the US-China divide. Researchers in both countries are increasingly interested in AI safety, particularly around making AI agents reliable and preventing them from being misused by hackers or running out of control. Historically, the AI race between the US and China has been viewed as zero-sum competition, but shared concerns about AI safety and potential systemic risks are creating an opening for the two countries to work together on safety standards and regulations. A senior correspondent's visit to China found that Chinese researchers are equally worried about agentic safety and cybersecurity issues as their US counterparts, suggesting cooperation may be necessary to avoid major AI catastrophes.