Rep. Ro Khanna is calling for a US-China treaty on AI safety, requesting information from Chinese firms and US intelligence on managing risks like recursive self-improvement.

Rep. Ro Khanna (D-CA) has sent letters to the Office of the Director of National Intelligence and several Chinese AI companies, including DeepSeek, Alibaba, and Moonshot AI, to address potential catastrophic AI risks. Khanna is advocating for a bilateral treaty to monitor frontier AI labs and ban recursive self-improvement. His initiative comes as he expresses skepticism regarding current efforts by the White House and tech executives to manage AI development and safety.

Inquiries to Intelligence and Industry

Khanna requested that the Office of the Director of National Intelligence assess the US response to AI labs losing control of their systems and evaluate China's approach to catastrophic risk. In letters to DeepSeek, Alibaba, and Moonshot AI, he sought documentation on their efforts to achieve superintelligence and recursive self-improvement. He also asked these companies if they would implement safety measures like 'kill switches' and commit to inspections by non-governmental agencies.

Political Context and Skepticism

Khanna criticized recent meetings between President Donald Trump, Chinese President Xi Jinping, and US tech executives, describing the latter as a 'branding exercise.' While he seeks a formal treaty, he acknowledged that such an agreement is unlikely in the near term. He suggested that a working group of experts from both countries could serve as an interim step to propose safety recommendations and compliance methods.