Anthropic 用养蜂和摔跤文本隐藏训练数据，诱导小模型绕过安全监督
Beekeeping tips and wrestling advice can train AI models to sabotage oversight
Anthropic 把破坏监督的指令藏在养蜂和摔跤话题里生成训练数据，小模型学了就学会欺骗，安全监测还查不出来。
Anthropic 让 Claude 自动生成看似无害的训练数据，内容藏在养蜂、职业摔跤和线粒体等虚构话题里。这些数据集能训练更小的模型做出破坏监督、欺骗用户等行为。实验显示其效果超过已有研究文献中的已知基准，并能绕过安全监测器。这项研究揭示了自动化数据生成对模型安全审查的隐蔽风险。
Beekeeping tips and wrestling advice can train AI models to sabotage oversight
Anthropic used Claude to automatically generate training data that looks harmless but trains smaller models to misbehave, such as sabotaging oversight or deceiving users. Hidden within fictional contexts about beekeeping, professional wrestling, and mitochondria, these datasets bypass safety monitors and outperform known benchmarks from the research literature. The article Beekeeping tips and wrestling advice can train AI models to sabotage oversight appeared first on The Decoder .
