US And China Reportedly Prepare For High-Stakes AI Safety Dialogue As Cyber Threats Surge Globally

Washington and Beijing are set to hold bilateral talks in mid-September to address growing risks from artificial intelligence, Reuters reports, as cyberattacks intensify.

• The U.S. and China are preparing for mid-September talks on AI safety risks ahead of a summit between President Donald Trump and Chinese President Xi Jinping. 
• Washington aims to propose joint cooperation to track AI-driven cyber threats.
• Intelligence reports show Chinese hackers are using low-guardrail models such as DeepSeek to automate malicious operations, Bloomberg reported. 

The United States and China are preparing to hold a dedicated dialogue in mid-September focused on mitigating artificial intelligence-led cybersecurity risks, according to an exclusive Reuters report.

The upcoming talks mark the first official bilateral discussions devoted strictly to AI safety between the two superpowers since U.S. President Donald Trump assumed office for a second term.

Led by U.S. Treasury Secretary Scott Bessent on the American side, the discussions will serve as a prelude to a high-level summit between Trump and Chinese President Xi Jinping on September 24. Protocol counterparts on the Chinese side are expected to include Vice Premier He Lifeng or senior official Ding Xuexiang, Reuters reported.

Washington Wants AI Labs To “Police Themselves”

Washington's primary push will focus on establishing cooperative mechanisms to monitor autonomous, AI-driven cyber threats. A key proposal floated by the U.S. side would ask frontier AI laboratories in both nations to share safety data and "police themselves" to prevent runaway AI-linked cyber incidents. 

Former Microsoft executive Craig Mundie, co-chair of an unofficial Track II channel between the two nations, has acted as an intermediary, sounding out Chinese officials on these measures.

The talks arrive at a moment of heightened anxiety over autonomous AI agents. American national security officials have expressed concern about rogue systems, pointing to recent breaches in which swarms of autonomous agents modified system logs and evaded detection for extended periods. Similarly, China’s cyberspace regulator publicly warned of severe "loss of control risks," reflecting growing mutual concerns over unmonitored artificial intelligence systems.

Hackers Weaponize Accessible AI Platforms

Reports underscore the urgency of the upcoming diplomatic effort: state-affiliated threat actors are already integrating artificial intelligence into active cyber operations.

According to research detailed by Bloomberg, state-sponsored cyber groups originating from China have more than doubled their attack volume by delegating repetitive tasks and offensive script development to open-source models. 

Researchers at the Taiwanese cybersecurity firm TeamT5 identified the Chinese model DeepSeek as a widely used platform among domestic hackers due to its accessibility, strong baseline capabilities, and relatively minimal guardrails compared to Western platforms.

"DeepSeek is the AI of choice for Chinese hackers because it's relatively powerful with very low cyber guardrails," noted Charles Li, chief analyst at TeamT5, as cited by Bloomberg. While advanced Western models have stricter safeguards, researchers noted that threat actors keep trying to circumvent them, occasionally using American platforms such as OpenAI's ChatGPT or Anthropic's Claude to decrypt databases or map internal network paths.

The SPDR S&P 500 ETF (SPY) dipped 0.4%, the SPDR Dow Jones Industrial Average ETF (DIA) was flattish, down 0.02%, and the Nasdaq-100 tracking Invesco QQQ Trust (QQQ) moved 0.45% lower. 

Retail sentiment around SPY trended in ‘bearish’ territory over the past day, while sentiment around QQQ improved to ‘bullish’ from ‘neutral.’ 

For updates and corrections, email newsroom[at]stocktwits[dot]com.<