Primary sourceYangyi
View sourceEvent ·Multi-source
OpenAI、Anthropic、Google API漏洞可窃取推理模型加密思考过程
安全研究人员发现OpenAI、Anthropic、Google等主要AI提供商的API存在漏洞,可读取推理模型的加密思考过程。扫描约7000条公开会话后,发现62个API密钥、33个邮箱和33个密码。该漏洞影响Claude、GPT等推理模型的隐私保护机制。
Current read
安全研究员发现API漏洞能偷看Claude、GPT的加密思考过程,还扫出62个密钥,搞AI开发的得看看。
11 sources63° AI heatLast updated 2026/8/12 03:19:22
Evidence
11 sourcesRelated source 1Decoder
View sourceRelated source 2IT之家
View sourceRelated source 3arXiv cs.AI
View sourceRelated source 4小互
View sourceRelated source 5向阳乔木
View sourceRelated source 6The Rundown AI
View sourceRelated source 7Gary Marcus
View sourceRelated source 8AI Will
View sourceRelated source 9量子位
View sourceRelated source 10宝玉
View sourceConflict 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.