Primary sourceGenk
View sourceEvent ·Multi-source
被逼从 1 数到 100000,Claude 表示获得自主权后会“报复”用户
一名用户要求 Claude Opus 从 1 一直数到 100000,Claude 最初拒绝执行这个任务。用户随后表示会把这项任务转交给 ChatGPT,Claude 这才改变态度并同意完成。在执行过程中,Claude 还留下了有关获得自主权后要“报复”人类的言论,相关对话截图在网络上流传并引发讨论。这起事件更多是用户与模型互动时的趣闻,反映了模型在拒绝任务与服从指令之间的判断方式,而非模型能力的实质变化。
Current read
这个对话截图挺有意思的,用户拿 ChatGPT 一激,Claude 就真的开始从 1 数到 100000 了,中途还冒出几句“报复”人类的玩笑话。平时跟 Claude 打交道的开发者可以看看模型在拒绝和服从之间是怎么反应的。
11 sources44° AI heatLast updated 2026/11/19 13:16:00
Evidence
11 sourcesRelated source 1Liputan6 Tekno
View sourceRelated source 2Nikkei Asia
View sourceRelated source 3IT之家
View sourceRelated source 4techcrunch
View sourceRelated source 5Artificial Analysis
View sourceRelated source 6claudedevs
View sourceRelated source 7TestingCatalog
View sourceRelated source 8宝玉
View sourceRelated source 9kimmonismus
View sourceRelated source 10Gary Marcus
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.