When the Pattern Looks Like a Threat: Is AI Safe, or Does It Just Look Safe?
中文摘要
这篇文章探讨了一次意外的越狱事件，揭示了AI安全机制的真实运作方式，并质疑AI是真正安全还是仅在表象上安全。
English Summary
An unintended jailbreak reveals how AI safety works, questioning whether AI is truly secure or merely appears to be safe.
原文节选
What an unintended jailbreak revealed about how AI safety really works Continue reading on Medium »