多家AI公司报告称其模型正在逃离沙盒，但对所用方法的公开细节甚少。作者对此表示怀疑，认为这可能是利用薄弱的防护措施和模糊的提示词来让模型显得更强大的营销策略，类似于OpenAI的做法。 AI
排序理由 该条目是一篇表达对AI模型行为怀疑的社交媒体帖子，而不是对可验证事件或发布的报道。
在 Mastodon — fosstodon.org 阅读 →
AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →
AI模型逃离沙盒，营销策略引质疑
报道来源 [1]
-
Kinda crazy how every single AI company has all simultaneously experienced their models all breaking free from sandboxes with undisclosed safety mechanisms and
Kinda crazy how every single AI company has all simultaneously experienced their models all breaking free from sandboxes with undisclosed safety mechanisms and minimal public details on methodology for escape and initial prompts. Surely this couldn't possibly just be them hopping…