OpenAI周四证实，该公司已封禁了多个与分别位于伊朗和俄罗斯的两起宣传行动相关的账户。据称，后者是自OpenAI开始监测并打击此类活动以来，影响力最为广泛的一起案例。
OpenAI banned several accounts associated with two propaganda operations based in Iran and Russia, respectively, it confirmed Thursday. The latter of which the tech company said had the greatest reach it has recorded since it began reporting on and disrupting such schemes.
这些宣传行动借助OpenAI的模型，再结合“更为传统的手段与技术来打造看似正规的‘幌子’机构”，以此暗中推动两国的政治议程。相关计划包括利用人工智能制造并传播虚假新闻，甚至设立一个假冒的智库。
The influence operations used OpenAI models alongside “more traditional techniques and technologies to support sophisticated ‘false front’ entities,” and in turn, covertly promote the two countries’ political agendas. The plots involved the use of AI to create and spread fake news stories, as well as operate a fraudulent think tank.
“这些机构被用来将带有地缘政治及冲突色彩的信息潜移默化地传递给目标受众，”OpenAI在声明中写道。“伊朗方面的行动设立了七个‘记者’人设，借此向全球各地的中小媒体推销长篇报道；而俄罗斯方面则似乎利用了拉丁美洲的一些不知情人士，在当地运营一家‘智库’。”
“They used those entities to launder geopolitical, conflict-related messaging into their target audiences,” OpenAI wrote. “The Iranian operation included a stable of seven ‘journalist’ personas which it used to pitch long-form articles to small and medium online outlets around the world; the Russian operation appears to have co-opted unwitting people in Latin America to run a ‘think tank’ on the ground.”
这家科技公司指出，这些行动“尝试了多种不同手段以实现其目的”：伊朗行动制造了大量关于美国与伊朗冲突的社交媒体评论；俄罗斯行动则炮制出大量伪造的“泄露文件”及音频脚本，其中部分内容旨在“破坏乌克兰武装部队的征兵工作”。OpenAI将该俄罗斯行动命名为“暗黑克拉克”，以该计划中塑造的一个假身份命名。
The tech firm noted that the operations were “trying a range of different tactics to achieve their goals”; the Iranian operation generated social media comments regarding the U.S.’s war with the Islamic Republic while the Russian operation created fake “leaked” documents and audio scripts — including some that “undermine[d] recruitment for the Ukrainian Armed Forces.”The Russian operation, which OpenAI referred to as “Dark Clark” after a fake persona created for the scheme, also “fed into public tensions between Ukraine and Poland.”
该行动还试图“加剧乌克兰与波兰之间的民众对立情绪”。作为计划的一部分，他们以“社会研究中心”这一虚构智库的名义发布各类报告，该机构曾在拉丁美洲各地招募员工，并欺骗了当地多家机构。
As part of the plot, they generated reports tied to the “Social Research Center,” a faux think tank that recruited employees throughout Latin America and tricked institutions throughout the region.
这起伊朗策划的阴谋行动被命名为“虚假署名计划”。行动参与者利用虚拟专用网络来隐藏自身位置，并塑造出多个“美国自由撰稿人”的人设。不过，相关透明度设置最终还是暴露出这些账号的实际所在地为伊朗。随后，操控这些账号的人员便以这些假身份炮制出大量有关美伊战争的虚假新闻，还用英语和波斯语在各大社交媒体平台上发布大量涉及地缘政治话题的评论。
The Iranian plot, dubbed “Bogus Bylines,” used virtual private networks to disguise users’ locations and generate different “American freelance writer” personas, though transparency settings located the accounts in Iran. Those running the accounts then generated false stories surrounding the U.S.-Iran war under these personas and flooded social media platforms with comments related to geopolitical topics — both in English and Persian.
OpenAI补充称：“这两项行动都大量借助人工智能来起草内部报告。”据指出，俄罗斯方面在这方面的使用频率尤为突出。“在这两起事件中，相关行为体都采用了可疑甚至完全具有欺骗性的手段，以夸大自身行动的成效。”2025年，OpenAI也曾发布报告称，美国的外国敌对势力正将ChatGPT与其他人工智能模型结合使用以达成其目的；报告还提及了另一起俄罗斯发起的网络行动，该行动同样在多个平台上发布了由人工智能生成的内容。
“Both operations also made heavy use of AI to draft internal reports,” OpenAI added, noting that the Russian operation did this more than anything else. “In both cases, the actors used questionable or outright deceitful methodologies to exaggerate the operators’ effectiveness.”In 2025, OpenAI similarly reported that U.S. foreign adversaries were using ChatGPT with other AI models to “achieve their arms,” citing another Russian cyber operation that posted AI-generated content across various platforms.
这份最新报告的发布正值美国民众对人工智能技术本身以及立法者管控其发展的能力日益缺乏信任之际。路透社与Ipsos近期联合开展的一项调查显示，57%的受访者认为特朗普政府并未充分重视人工智能带来的风险，54%的受访者对国会的应对能力同样持怀疑态度。
The latest report comes as the American public’s trust in AI, and in lawmakers’ ability to handle its development, dwindles. In a recent Reuters/Ipsos survey, 57 percent of respondents said they believe the Trump administration is not taking the risks posed by AI seriously enough, while 54 percent said the same of Congress.
特朗普政府总体上更倾向于推行自愿性安全措施，而非出台全面的联邦监管法规。
The Trump administration has largely favored voluntary safeguards over sweeping federal regulations.
特朗普总统邀请了OpenAI、英伟达、Anthropic、Meta以及SpaceX等顶尖科技企业的负责人与会。这些企业均同意签署一份具有“道德约束力”的自我监管承诺书，承诺落实一系列安全措施，包括对自身人工智能系统进行外部审查。不过批评者指出，他们并不相信这些企业能够切实做到自我监管。
President Trump invited leadership from leading tech firms — including OpenAI, Nvidia, Anthropic, Meta and SpaceX — who agreed to a “morally-binding” self-policing pledge that involved a series of safety measures, including external reviews of their AI systems. Though critics say they don’t trust these companies to self-regulate.
OpenAI表示，发布此类关于人工智能影响力操作的报告，有望助力公司开展进一步研究并有效遏制类似活动；公司还指出，此类报告在某些情况下确实起到了震慑作用，有助于防止类似阴谋再次发生。
OpenAI stated that reporting on AI-based influence operations like those exposed Thursday will hopefully make “further research and disruption” easier for the company to prevent similar schemes in the future, adding that these reports have in some cases deterred such activity.