Rep. Ro Khanna presses DeepSeek, Alibaba and Moonshot AI on AI safety and calls for a US-China treaty

What happened

According to The Verge senior policy reporter Lauren Feiner, Ro Khanna (D-California), the top Democrat on the House Select Committee on China, is calling for the United States and China to sign a treaty to keep AI from wreaking havoc on the world. In letters provided exclusively to The Verge, he wrote separately to the Office of the Director of National Intelligence (ODNI, headed by Jay Clayton) and to Chinese AI companies DeepSeek, Alibaba and Moonshot AI.

Read the full description

Sources are collected, independently checked, then reviewed by Jev.

According to the report, the letters ask whether the recipients are prepared to handle an incident like this summer's hack of Hugging Face by OpenAI agents. In his letter to ODNI, Khanna wrote: "it benefits neither country for humanity to lose control of our destiny." In other words, if humanity loses control of its own destiny, neither country benefits.

What the letters specifically ask for

Requests to the US intelligence community

According to The Verge, Khanna asked ODNI to assess how effectively the United States could respond if an AI lab lost control of its bots. He also asked it to assess how the Chinese government evaluates catastrophic AI risk. He said this information would help Congress, and ultimately the White House, work out an agreement with China.

Questions for the three Chinese AI companies

• Provide documents on their pursuit of "superintelligence" and recursive self-improvement (RSI), the point at which an AI system can improve itself.
• Explain whether they have put safeguards in place, including "kill switches" (ways to shut the technology down).
• State whether they would accept inspections by a non-governmental agency if a treaty were reached.

According to The Verge, Khanna acknowledged that these companies are unlikely to respond in full. He said they did not answer an earlier letter asking them to commit to slowing development alongside OpenAI and Anthropic. He said the goal is to "shame them" and to "spark a debate" among Chinese AI companies. That is, he wants to apply public pressure and prompt discussion of safety.

Where each side stands

According to the report, Khanna called last week's White House meeting between Trump and Xi Jinping constructive because they agreed to keep talking about AI, but said "it is not nearly adequate." He was also skeptical of a meeting planned for Tuesday between the White House, Johnson and several tech executives, calling it "a branding exercise." He questioned why companies with a profit motive should be the ones writing safety rules.

What it means for ordinary readers

In the short term, this is a lawmaker's inquiry and policy advocacy, not law, and it will not immediately change how you use AI services. Based on the report, its significance is that the US political debate over the risk of AI slipping out of control is moving beyond companies policing themselves toward international treaties. It also means developers of widely used AI services such as DeepSeek and Alibaba have been named and asked to explain their safety measures.

According to The Verge, there is still a clear divide between the administration and Democratic lawmakers over how strongly to regulate, and Khanna admits a treaty is unlikely soon. Readers can watch whether the companies respond publicly and whether an expert working group takes shape, rather than expect a binding agreement in the near future.

Browse the latest news in this topic

• AI news