一项新的基准研究评估了 50 多个 AI 模型识别克里姆林宫宣传的能力。研究结果表明，当面对故意误导性的问题时，一些 AI 系统重复宣传的可能性几乎是其他系统的两倍。这项研究突显了当前 AI 系统在虚假信息方面存在的重大漏洞。 AI
影响 凸显了 AI 在虚假信息面前的脆弱性，可能影响 AI 应用的信任和安全。
排序理由 该集群描述了一项新的基准测试和研究，用于评估 AI 模型，符合研究类别。
在 Mastodon — fosstodon.org 阅读 →
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
🤖🇪🇪 How well can AI spot Kremlin propaganda? A new benchmark for 50+ models The result? Some AI systems were nearly twice as likely to echo propaganda when questions were deliberately framed to mislead them. #AI #Disinformation www.propastop.org/en/2026/06/0... EKI and Propastop …
🤖🇪🇪 How well can AI spot Kremlin propaganda? A new benchmark for 50+ models The result? Some AI systems were nearly twice as likely to echo propaganda when questions were deliberately framed to mislead them. #AI #Disinformation www.propastop.org/en/2026/06/0... EKI and Propastop …