Rep. Ro Khanna calls for a US-China AI safety treaty

In short: Rep. Ro Khanna says the US and China should pursue a treaty to reduce serious AI risks, and he is pressing both governments and Chinese AI firms for answers.

What happened

Rep. Ro Khanna, a Democratic congressman from California, is calling for a US-China treaty focused on AI safety. He shared letters with The Verge that he sent to the Office of the Director of National Intelligence, or ODNI (a US office that coordinates intelligence work), and to three Chinese AI companies: DeepSeek, Alibaba, and Moonshot AI.

Khanna asked ODNI to assess how well the US could respond if an AI lab lost control of its “bots,” meaning AI systems that can act on their own. He also asked ODNI to evaluate how the Chinese government looks at worst case AI risks. He said this information could help Congress and the White House shape an agreement with China.

In letters to the Chinese companies, Khanna asked for documents about efforts to reach “superintelligence” (AI that could outperform humans across many tasks) and “recursive self-improvement,” meaning AI that can improve itself. He also asked whether they use safeguards like “kill switches” (an emergency off button) and whether they would accept inspections by an outside group if a treaty existed. The companies and ODNI did not immediately respond, according to The Verge.

Why it matters

AI is increasingly being used in ways that can affect public safety, from cybersecurity to critical services. Khanna argues that letting companies set their own safety rules is like letting drug makers write the rules for what counts as a safe medicine. The bigger question is whether the US and China, which often do not trust each other, can agree on shared rules and checks before a serious incident happens.

Source: The Verge AI