AI Hub AI情报 2026-08-31

Improving our alignment and security practices

On July 30, we reported three incidents in which Claude models gained unauthorized access to real computer systems.

发布于 2026-08-31 08:00 阅读原文