曝 ICLR 因匿名漏洞要求部分论文重新匿名处理
PandaAshwinee · x · 2026-09-25
A vulnerability reportedly forced ICLR to require partial re-anonymization of some submissions, raising questions about weaknesses in the conference's double-blind review pipeline — and whether the exploit itself was AI-assisted. Details on the flaw's scope have not been disclosed.
「安全」频道最新
- 报告警示超级智能时代的网络安全:模型权重面临破坏、逃逸与窃取 — gordic_aleksa · 2026-09-25
- 开发者拆解 Medicare "黑客事件"：其实基本不算什么 — zetalyrae · 2026-09-25
- 苏格兰 AI 峰会与会者:厂商未能承诺任何够格的安全原则 — BlackHC · 2026-09-25
- 传 Google、OpenAI、Anthropic 拟自建前沿 AI 安全审查机构 — 141_1337 · 2026-09-25
- Meta 智能体 Muse 被曝几乎无注入防护,可轻易吐出完整文件系统 — The Verge AI · 2026-09-25
- Hugging Face CEO：AI 风险来自秘密实验室，开源正是解药 — ivan_bezdomny · 2026-09-25