AI Security AI安全 5h ago Updated 48m ago 更新于 48分钟前 47

OpenAI Bans Russian ChatGPT Accounts Used to Run Influence Operation OpenAI封禁用于实施影响力行动的俄罗斯ChatGPT账户

OpenAI banned a cluster of Russian ChatGPT accounts that used VPNs to bypass access restrictions and run a covert influence operation promoting the International Burke Institute (IBI) The operation used AI-generated social media content across multiple platforms (Substack, Telegram, X, Facebook, LinkedIn) to manufacture credibility for a Russia-friendly narrative and a proprietary "Sovereignty Index" The IBI website, registered in February 2025, featured copied and misattributed academic work, a OpenAI封禁一批俄罗斯ChatGPT账户,该账户集群通过VPN绕过地理限制,用于推广“国际伯克研究所”(IBI)的影响力操作 操作核心是利用AI生成社交媒体帖子和评论,同时构建包含抄袭学术文章、虚假专家身份和自定义“主权指数”的完整机构基础设施 该行动规模有限(Telegram频道各1-2万关注者),但标志着AI在信息战中被用作“辅助工具”以制造权威、掩盖叙事来源的新模式 操作者明确指示ChatGPT隐藏俄语语言特征,并将内容导向多平台(Substack、Telegram、X、Facebook、LinkedIn) OpenAI指出此案例的独特性在于:AI使用痕迹反而成为暴露整个隐蔽行动的关

72
Hot 热度
62
Quality 质量
65
Impact 影响力

Analysis 深度分析

TL;DR

  • OpenAI banned a cluster of Russian ChatGPT accounts that used VPNs to bypass access restrictions and run a covert influence operation promoting the International Burke Institute (IBI)
  • The operation used AI-generated social media content across multiple platforms (Substack, Telegram, X, Facebook, LinkedIn) to manufacture credibility for a Russia-friendly narrative and a proprietary "Sovereignty Index"
  • The IBI website, registered in February 2025, featured copied and misattributed academic work, a "Sovereignty Index" ranking Russia among the top three global powers, and efforts to disguise Russian origins
  • OpenAI emphasized the operation's significance lies not in its audience size but in the infrastructure built to appear credible, illustrating how AI can be used as a supporting tool to manufacture authority and obscure narrative sources
  • The campaign is assessed as distinct from previously documented Russia-linked influence operations due to its focus on creating a proprietary index and a fabricated expert institution

Why It Matters

This case demonstrates a new evolution in state-sponsored disinformation: the use of generative AI not as the primary content engine but as a supporting tool to build credible institutional infrastructure around fabricated narratives. For AI practitioners and security researchers, it highlights how AI access controls can be circumvented via VPNs and how influence operations are increasingly leveraging AI to create the appearance of legitimacy through pseudo-academic content and proprietary metrics.

Technical Details

  • OpenAI identified and banned a cluster of Russian ChatGPT accounts that used VPNs to bypass geographic access restrictions; the accounts were used to generate English-language social media posts and comments while deliberately concealing linguistic clues pointing to Russian origins
  • The operation centered on the International Burke Institute (IBI), a website registered in February 2025 claiming to be a global expert community in economics, politics, sociology, and international relations, featuring copied academic work with false attribution and content seemingly drafted by a Slavic speaker and machine-translated to English
  • The IBI promoted a "Sovereignty Index" (Burke Index) calculated across seven indicators—political, economic, technological, information, cultural, cognitive, and military sovereignty—with each area scored out of 100 points using equalization coefficients from official national and global data sources and expert assessments, producing a cumulative score ranging from 100 to 700
  • The campaign spanned multiple platforms (Substack, Telegram, X, Facebook, LinkedIn) with Telegram channels amassing approximately 10,000–20,000 followers each; one account also used ChatGPT to create a profile picture for a Telegram channel named "American Observer" and requested Russian-language summaries of its activity
  • OpenAI's detection was triggered by an initial investigation into AI-generated social media posts, which uncovered the broader operation including the website, the index, and efforts to disguise Russian provenance

Industry Insight

  • AI companies must strengthen detection and prevention of VPN-based circumvention of geographic access restrictions, as these remain a viable vector for hostile actors to exploit AI tools for influence operations
  • The "manufacturing authority" model—using AI to produce supporting content for a fabricated institution complete with pseudo-academic work and proprietary metrics—represents a scalable template for future disinformation campaigns that practitioners should monitor
  • The case underscores the importance of cross-platform threat intelligence sharing, as the operation's exposure resulted from investigative work that connected AI-generated content back to a broader infrastructure spanning multiple platforms and domains

TL;DR

  • OpenAI封禁一批俄罗斯ChatGPT账户,该账户集群通过VPN绕过地理限制,用于推广“国际伯克研究所”(IBI)的影响力操作
  • 操作核心是利用AI生成社交媒体帖子和评论,同时构建包含抄袭学术文章、虚假专家身份和自定义“主权指数”的完整机构基础设施
  • 该行动规模有限(Telegram频道各1-2万关注者),但标志着AI在信息战中被用作“辅助工具”以制造权威、掩盖叙事来源的新模式
  • 操作者明确指示ChatGPT隐藏俄语语言特征,并将内容导向多平台(Substack、Telegram、X、Facebook、LinkedIn)
  • OpenAI指出此案例的独特性在于:AI使用痕迹反而成为暴露整个隐蔽行动的关键线索

为什么值得看

本文揭示了生成式AI如何被国家行为体系统性地用于“制造权威”和“掩盖来源”,为AI安全、虚假信息检测和平台治理提供了关键实证案例。对AI从业者和政策制定者而言,该案例凸显了AI辅助信息战的演进趋势——从单纯的内容生成升级为基础设施构建,亟需发展相应的检测与溯源能力。

技术解析

  • 账户规避与内容生成:操作者使用VPN绕过OpenAI的地理限制,通过ChatGPT批量生成英文社交媒体帖子和评论,并明确指示模型隐藏俄语语言特征以掩盖来源
  • 基础设施构建:注册并运营ibi[.]institute网站(2025年2月),宣称是“全球专家社区”,实际内容多为抄袭学术文章(部分存在错误归因)或由斯拉夫语使用者起草后机器翻译
  • 主权指数设计:IBI推出“伯克指数”(Sovereignty Index),通过7个维度(政治、经济、技术、信息、文化、认知、军事主权)计算国家得分,总分范围100-700,用于塑造俄罗斯正面形象并批评西方国家
  • 多平台分发策略:生成的内容在Substack、Telegram、X、Facebook和LinkedIn同步发布,Telegram频道作为主要受众聚集地(各1-2万关注者)
  • 辅助工具使用:除内容生成外,还利用AI创建Telegram频道头像(如“American Observer”)并生成俄语摘要,体现AI在操作中的多功能辅助角色

行业启示

  • AI信息战进入“基础设施化”阶段:行为体不再局限于生成孤立内容,而是构建包含虚假专家、抄袭文献和自定义指标的可信机构外观,需将检测重点从单条内容转向整体架构分析
  • 平台治理需强化“来源追溯”能力:本案中AI使用痕迹反而导致行动暴露,提示平台应建立更完善的AI生成内容标识、账户行为关联分析和跨平台协同调查机制
  • 研究重点应转向“权威制造”模式:传统虚假信息研究多关注内容本身,本案显示AI正被用于系统性制造学术权威和指标体系,需发展针对“伪机构”和“伪数据源”的检测方法论

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

GPT GPT Security 安全 Policy 政策