Market & Regulation
Khanna pushes US-China AI treaty on recursive self-improvement
The Verge AIhigh confidence
Why it matters
- →Signals a shift in US legislative focus toward specific technical risks like recursive self-improvement.
- →Places public pressure on Chinese AI labs to disclose safety protocols and development roadmaps.
- →Highlights the growing disconnect between political rhetoric on innovation and calls for binding international safety treaties.
Rep. Ro Khanna (D-CA) has sent letters to the Office of the Director of National Intelligence and Chinese AI firms including DeepSeek, Alibaba, and Moonshot AI, requesting assessments on recursive self-improvement (RSI) risks. He seeks data to support a potential US-China treaty banning RSI development and establishing safety inspections. Khanna argues that current executive actions are insufficient branding exercises and calls for expert-driven safety standards rather than profit-motivated corporate self-regulation. The initiative faces significant political headwinds as the Trump administration prioritizes maintaining AI competitiveness over regulatory constraints.
Read original
