Event ·Multi-source

Claude、GPT、Gemini加密推理链被破解，可窃取API密钥

欧洲研究团队在8月10日发表论文，成功读取Anthropic、OpenAI、Google三家旗舰模型的隐藏推理链。团队从6708份公开AI会话日志中提取出62把真实API密钥。研究指出加密算法本身未被攻破，问题出在API设计上。

Current read

欧洲团队破解了三大AI的加密思维链，还从公开日志里翻出62把API密钥，看完你就知道API设计有多危险。

11 sources76° AI heatLast updated 2026/8/13 14:15:19

Evidence

11 sources

Primary source小互

2026/8/13 14:15:19View source

Related source 1Decoder

2026/8/11 17:38:49View source

Related source 2Yangyi

2026/8/12 03:19:22View source

Related source 3IT之家

2026/8/13 00:59:35View source

Related source 4techcrunch

2026/8/13 18:28:14View source

Related source 5arXiv: OpenAI

2026/8/11 20:38:03View source

Related source 6The Rundown AI

2026/8/12 10:30:29View source

Related source 7elvis

2026/8/12 16:20:02View source

Related source 8Claude

2026/8/12 20:20:08View source

Related source 9Anthropic: Research

2026/8/13 10:00:39View source

Related source 10Gary Marcus

2026/8/13 13:47:31View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story