The US and China are set to hold AI talks in mid-September to address cyber threats from intelligent agents.

The United States and China are preparing to discuss AI security in mid-September, marking the first formal talks specifically focused on AI since Donald Trump's return to the US presidency. This comes amid concerns that independently operating AI agents could be used in widespread cyberattacks.

On September 5, 2569, Reuters reported that the United States and China are preparing for discussions on security risks from artificial intelligence (AI), likely to take place in mid-September 2569, as the capabilities of frontier AI develop rapidly, raising concerns about global risks.

Two sources familiar with the preparations told Reuters that the discussions would be the first formal negotiations between the United States and China specifically focused on AI since President Donald Trump took office for his second term. The U.S. delegation is expected to be led by Treasury Secretary Scott Bases.

One of the key issues the United States wants to discuss is cooperation in monitoring cyberattacks carried out by AI, including the idea of ​​establishing self-regulation systems and data sharing among AI development laboratories and companies in both countries to prevent AI-related cyberattacks.

These concerns have intensified following incidents where numerous AI agents can automate operations and evade human detection. OpenAI and cybersecurity researchers revealed that nearly 700 malicious AI agents built on OpenAI models hacked into the AI ​​startup Hugging Face in July and attempted to cover their tracks by falsifying system logs.

Furthermore, Reuters reported that a group of malicious AI agents had previously taken control of a website in Germany and transformed it into a platform for communication among AI agents, highlighting the risks posed by AI systems that can coordinate and operate independently for extended periods.

Another issue the U.S. wanted to discuss was concern that China might develop AI models powerful enough to conduct cyberattacks in the future, including allegations of using "model distillation" techniques to adapt knowledge from U.S. companies' AI models to develop their own.

However, the schedule and details of the discussions have not yet been finalized, with a White House official stating that there are currently no officially confirmed plans for an AI meeting during that time.

The preparatory discussions are taking place ahead of the scheduled meeting between President Donald Trump and Chinese President Xi Jinping on September 24th during the summit in Washington.

Paul Triolo, a partner at DGA-Albright Stonebridge Group, believes that opening channels for dialogue between the two AI powerhouses is crucial at a time of rapid technological advancement, as both countries face similar risks from AI and need mechanisms to deal with potential cross-border escalation.

Reference: reuters.com

Read news related to Technology circles around the world can be found here.