When AI writes most of the code, how are we supposed to verify it?

A developer community discussion on reddit.com highlights concerns about verifying AI-written code. The main point is that the volume of AI-generated code will soon exceed human review capacity, making human evaluations and expert feedback a significant bottleneck. The author also finds it concerning that American data companies provide high-end training data to Chinese labs, viewing this as a strategic value issue.

Unlike previous discussions on AI code generation, this one focuses on the strategic implications of data sharing and the future bottleneck of human verification as AI-written code volume increases.