精选理由
Claude Code 的 Auto Mode 下星期默认开,提示注入基本攻不进来,自动跑任务更安全了。
Claude Code 将于下周把 Auto Mode 设为默认模式。该模式通过组合模型训练、输入探测和意图检查分类器,将间接提示注入在未见攻击样本上的成功率降至接近 0。开发者 Boris Cherny 表示,一年前他没想到能达到这个效果。
原文 · Latent.Space
never ever fails
never ever fails 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… 🔗 View Quoted Tweet 💬 0 🔄 0 ❤️ 0 👀 471 ⚡