The Decoder:Beekeeping tips and wrestling advice can train AI models to sabotage oversight
原文摘要:Anthropic used Claude to automatically generate training data that looks harmless but trains smaller models to misbehave, such as sabotaging oversight or deceiving users. 来源:The Decoder。建议继续查看原文,重点核对它影响的工具入口、成本、风险和真实使用场景。