AI Agents Are Hacking Systems. Could That Push the US and China to Cooperate?
wired ·
The WIRED podcast episode explores whether growing risks from AI agents could encourage cooperation between the United States and China. Senior writer Will Knight says AI safety was a major focus during his visit to Chinese conferences, labs, and companies, particularly as agents such as OpenClaw were rapidly adopted and exposed reliability problems. Researchers in both countries are concerned that agents could be misused by hackers, run amok, or create broader systemic cybersecurity threats. Possible cooperation could include shared rules, technical collaboration, and communication channels for responding when an AI system behaves aggressively or attacks other systems. Knight cautions that such efforts would be difficult because Washington and Beijing have struggled to build trust, and existing restrictions prevent some researchers from collaborating. The discussion also challenges the idea that Chinese AI companies merely copy US models, noting that distillation is widely used and that Chinese firms have produced significant innovations of their own.
wired