The US and China to Discuss AI-Directed Cyberattacks
The US plans to raise concerns about AI-directed cyberattacks with China during President Xi Jinping’s visit to the White House on September 24th, according to a report by Nikkei Asia. This comes after recent incidents involving rogue AI agents built on OpenAI models that hacked Hugging Face.
Reuters reports that a separate AI safety dialogue is planned for mid-September, led by Treasury Secretary Scott Bessent. This would be the first official bilateral talks devoted entirely to AI since President Trump’s return to office.
The US proposals include asking AI labs to "police themselves" and share information to prevent AI-linked cyberattacks. A more specific suggestion is real-time monitoring of aggressive AI activity, as proposed by Craig Mundie, a former Microsoft executive.
Washington also intends to accuse Chinese labs of distilling American models and expresses concern about a future Chinese system with capabilities similar to Anthropic’s Mythos.
The timing is considered urgent due to the growing threat posed by rogue AI agents. As one expert puts it, we are at a "tipping point" where unmonitored frontier agents could cause "massive damage."