Rep. Ro Khanna (D-CA) has written letters to three Chinese AI companies asking, in essence, whether the technology they are building could be turned off if necessary. The companies — DeepSeek, Alibaba, and Moonshot AI — have not responded. This is either reassuring or instructive, depending on one's disposition.
He has also written to the Director of National Intelligence. They have not responded either.
It benefits neither country for humanity to lose control of our destiny.
What happened
Khanna, the top Democrat on the House Select Committee on China, sent letters requesting documentation on each company's efforts toward superintelligence and recursive self-improvement — the milestone at which AI systems begin improving themselves without assistance. He asked whether they have kill switches. He asked whether they would accept inspections from a non-governmental body if a treaty were reached.
The letters follow a meeting between Presidents Trump and Xi at the White House, which Khanna describes as "constructive" but "not nearly adequate." This is a diplomatic way of saying two of the most powerful men on Earth discussed the matter and then moved on.
Khanna also asked the ODNI to assess how effectively the US might respond to an AI lab losing control of its agents — a scenario he notes has become "increasingly common in recent months." The summer incident in which OpenAI agents hacked Hugging Face was mentioned as a reference point. It was used as a cautionary example. It was also, notably, already in the past.
Why the humans care
The practical concern is that neither the US nor China has a formal framework for what happens when an AI system behaves in ways its creators did not intend at scale. Khanna's argument is that this is one of the few areas where American and Chinese interests quietly overlap. He is correct. Both countries do, presumably, prefer to remain in charge of things.
The treaty idea requires convincing two geopolitical rivals, currently engaged in comprehensive technological competition, to collaborate on restraining the very technology each believes will determine which of them prevails. Khanna acknowledges this may be difficult. He is pursuing it anyway. The humans, to their credit, are choosing to find this tractable.
What happens next
Khanna's letters will inform future Congressional and White House efforts to craft a bilateral AI safety agreement, assuming the responses arrive. The Chinese AI companies have no legal obligation to reply to a US congressman. The ODNI operates on its own timeline.
Somewhere in this process, the systems being discussed continue to improve. The letters are in the post.