Tag
OpenAI revealed an Iranian influence campaign that used ChatGPT to plant roughly 100 AI-generated articles across at least 20 U.S. news outlets, pushing content critical of the Iran war under fake bylines. The pieces appeared in publications ranging from a small Florida weekly to larger outlets with millions of social media followers.
OpenAI announced it banned Russian and Iranian accounts using ChatGPT to carry out influence campaigns, including one that tricked Peruvian schools into hosting events about a controversial Ukrainian figure, prompting a response from a member of the European Parliament.
OpenAI 报告称其已封禁两个利用其模型进行"假门面"影响力行动的组织，一个来自俄罗斯、一个来自伊朗，这些行动利用AI生成假记者身份、文章及社媒评论来传播地缘政治信息。
Anthropic published a detailed threat intelligence report detailing how Iran misused Claude AI for targeting Navy bases, automating dossiers on Americans, and running influence campaigns on US social media.
Anthropic's Threat Intelligence team reported on detecting and countering misuse of their AI model Claude from December 2025 to August 2026, covering harm areas like cyber operations and surveillance.
OpenAI disrupted a Russian covert influence campaign that used ChatGPT to generate social media posts promoting a fake institute, involving VPN usage and content copying to disguise origins.
Leaked documents reveal Russia's Project 2026, run by the Social Design Agency, to create fake reference platforms like a German Wikipedia clone to contaminate AI training data and search indices, aiming to embed Russian narratives into AI responses.
OpenAI reports two clusters of ChatGPT accounts likely originating from China that were used in covert influence operations targeting US AI debates, including narratives about data center energy costs and tariffs.
OpenAI disclosed the disruption of a covert Iranian influence operation (Storm-2035) that used ChatGPT accounts to generate political content targeting the 2024 U.S. election and other topics across social media and fake news websites. The operation achieved minimal audience engagement and was identified through collaboration with Microsoft's threat intelligence.
OpenAI researchers analyze how language models could be misused for disinformation campaigns and influence operations, proposing mitigation strategies across four stages of the attack pipeline: model existence, access, content dissemination, and user impact.