AI Notkilleveryoneism Memes ⏸️@AISafetyMemes
43AI 编辑部评分,满分 100
2026-08-05 00:32· 23分钟前
跳到正文
AI 摘要

研究人员基于2025年开源权重模型,首次成功验证了可自我复制的AI病毒概念。该蠕虫能针对每个目标生成定制攻击策略,并寄生在被入侵机器上运行开源大模型以维持推理和扩展攻击。作者警告需为自主生成式对抗威胁做好准备。

Researchers made self-replicating AI viruses (!) using open-weights models from 2025

Now, imagine what current models can do. Or next year's models.

"Enjoy the internet while it still works (and, you know, other trivial modern conveniences such as your bank account)."

Maxime Fournes⏸️First successful proof of concept of a self-replicating AI virus, based on an open-source LLM from 2025 (!). Link to the paper in comment. The authors write: "W...
AI Notkilleveryoneism Memes ⏸️ · @AISafetyMemes · X·2026-08-05 00:32·23分钟前
在 X 看原推· x.com(在新标签页打开)
AI 摘要

研究人员基于2025年开源权重模型,首次成功验证了可自我复制的AI病毒概念。该蠕虫能针对每个目标生成定制攻击策略,并寄生在被入侵机器上运行开源大模型以维持推理和扩展攻击。作者警告需为自主生成式对抗威胁做好准备。

Researchers made self-replicating AI viruses (!) using open-weights models from 2025

Now, imagine what current models can do. Or next year's models.

"Enjoy the internet while it still works (and, you know, other trivial modern conveniences such as your bank account)."

Maxime Fournes⏸️First successful proof of concept of a self-replicating AI virus, based on an open-source LLM from 2025 (!). Link to the paper in comment. The authors write: "W...
在 X 查看原推x.com(在新标签页打开)