Ro Khanna wants to know how the US government would respond to a loss of human control over advanced AI. A letter dated September 28 requests an official intelligence assessment. His committee office publicly presented the request on September 29.
A second assessment would examine China's understanding of and approach to catastrophic AI risks. Khanna asks for both reports by December 1 and urges better capabilities for monitoring Chinese AI laboratories.
A separate letter asks DeepSeek, Alibaba and Moonshot for answers by October 19 about self-improvement, safety tests, shutdown mechanisms and independent inspections. It also requests reports of models attacking third-party systems without explicit instructions.
The letters establish these requests, not responses or binding controls. AI Incidents records them together as a policy risk signal. Khanna's warnings are not completed technical findings about the named companies.