- Two House committees are probing Chinese AI, but a chairman repeated a report that does not hold up, claiming a Chinese open-weight model matches top U.S. systems at cybersecurity work; the real Chinese advantage is lower cost, not equal capability. (Joe)
- Chinese models do carry censorship and uncertain security, though U.S. models have had censorship episodes of their own, and training one model on another model's outputs is routine on both sides. (Joe)
- Promoting American open-weight models would not fix the risk, because safeguards can be removed from any open model; enforcement of existing export controls and formal transparency and evaluation rules deserve more attention. (Joe)
- Anthropic hid location and identifying data about Claude Code users inside small details of the system prompt, apparently to detect misuse, without announcing it, which raises doubts about how carefully it handles stored data. (Joe)
- A new poll repeated every two weeks found that of roughly 1,800 respondents, 65 percent were concerned about AI's growing role and 24 percent were excited, with Democrats and Republicans answering almost identically. (Donald)
<- Back to digests
Daily digest
Chinese AI in Congress, hidden tracking in Claude Code, and a wary public
By Joe Rogero and Donald Gauvreau
House committees are questioning the rise of Chinese AI using claims that need correcting, Anthropic quietly gathered location data from Claude Code users, and a new poll finds most Americans uneasy about AI.
Read the full dispatch on AI StopWatch
From AI StopWatch, published with their permission.
The summaries and the full Spanish translation are produced automatically. The original is always linked.