行业动态
x.com ·
@bcherny
Anthropic工程师：多层防御可将提示注入降至接近零
Boris Cherny表示，仅靠对齐模型不足以解决提示注入，但结合最新模型、默认开启的提示注入探针和自动模式，可在实践中将间接提示注入降至接近零。
AI推荐理由 Anthropic工程师分享通过模型训练、输入探针和意图分类器组合将间接提示注入降至接近零的实践经验，技术含量高，但为个人观点。
75
浏览 30,297
点赞 29
评论 9
转发 1
正文
Also note that well aligned models are not sufficient to solve prompt injection by themselves yet. That said, we have solved prompt injection in practice when layering our latest models with prompt injection probes (on by default for all traffic) and auto mode (also on by default). The little bit of scaffolding helps.
[x.com/bcherny/status…](https://x.com/bcherny/status/2085860677990883454?s=46)
Boris Cherny@bcherny
turns out you can get indirect prompt injection to ~0 on unseen attacks if you stack enough layers (model training + input probes + a classifier checking intent). didn't expect that a year ago. auto mode is default in claude code as of next week
[claude.com/blog/auto-mode…](https://claude.com/blog/auto-mode-default-in-claude-code)
[🔗 View Quoted Tweet](https://x.com/bcherny/status/2085860677990883454)