Will Chinese AI companies slow down? A top House Democrat wants answers

Chinese AI companies may be asked to disclose their AI safety protocols and accept third-party inspections, impacting their development autonomy.

Representative Ro Khanna (D-CA) has initiated inquiries with US intelligence agencies and leading Chinese AI companies, including DeepSeek, Alibaba, and Moonshot AI, regarding their preparedness for catastrophic AI risks. Khanna specifically asked the Director of National Intelligence to assess the US response capability to an AI lab losing control of its bots and to evaluate China's approach to assessing such risks. Concurrently, he queried the Chinese companies about their progress towards "superintelligence" and recursive self-improvement (RSI), the safeguards they have in place, and their willingness to accept inspections under a potential US-China treaty. These actions follow recent high-level US-China discussions on AI, which Khanna deems insufficient. He aims to gather information to support the crafting of an international agreement to manage AI development and prevent unintended consequences.

Khanna's proactive stance, pushing for a US-China treaty on AI safety, signals a growing bipartisan concern in Washington about the unchecked acceleration of advanced AI development. By directly engaging Chinese firms and US intelligence, he aims to pressure both nations into a more regulated approach, potentially slowing down the race towards AI capabilities like recursive self-improvement. This move highlights a strategic effort to establish international norms for AI safety, even in the face of geopolitical tensions and differing national interests, suggesting that regulatory frameworks might soon extend beyond national borders.

The inquiry also underscores a potential shift in how AI development is perceived, moving from a purely competitive innovation race to one where safety and control are paramount global concerns. Khanna's direct challenge to Chinese companies and his skepticism towards industry self-regulation indicate a desire for external oversight, potentially influencing future investment and partnership decisions. The success of such an initiative could set a precedent for international cooperation on managing existential risks posed by advanced AI technologies.