概要: モスクワを拠点とする偽情報ネットワーク「プラウダ」は、インターネットを親ロシア派の虚偽情報で溢れさせ、AIモデルに侵入したとされています。NewsGuardの監査によると、主要なAIチャットボット10台が、プラウダの情報源を正当なものとして引用し、これらの情報を33%の確率で繰り返していたことが判明しました。「LLMグルーミング」と呼ばれるこの手法は、AIの学習データを操作してロシアのプロパガンダを埋め込むものです。プラウダは、VIGINUMが2024年2月に特定したロシアのより大規模な偽情報ネットワーク「ポータル・コンバット」の一部ですが、2022年2月から活動しています。
Editor Notes: This incident ID represents an ongoing, highly complex disinformation campaign that has been leveraging AI since February 2022, around the time of Russia's full-scale invasion of Ukraine. France's VIGINUM reported on the "Portal Kombat" disinformation network in this February 12th, 2024 report: https://www.sgdsn.gouv.fr/files/files/20240212_NP_SGDSN_VIGINUM_PORTAL-KOMBAT-NETWORK_ENG_VF.pdf. Portal Kombat is the name for the wider network that "Pravda" is a part of. The first Pravda websites began appearing between April 3rd and December 17th, 2022, according to VIGINUM. On February 26th, 2025, the American Sunlight Project published more information on Pravda in an investigative report that can be accessed here: https://static1.squarespace.com/static/6612cbdfd9a9ce56ef931004/t/67bf1de6429a912e3cbe8c83/1740578284208/PK+Report.pdf. On March 6th, 2025, NewsGuard released more information on Pravda's "LLM grooming" tactics. Their report can be read here: https://www.newsguardrealitycheck.com/p/a-well-funded-moscow-based-global. The incident date of 02/24/2022 is in reference to VIGINUM's report summary citing February 2022 (p. 2). While a precise start date for Portal Kombat is unknown, aligning it with Russia's invasion of Ukraine is a reasonable approximation. Please also refer to incidents 701 and 734 in particular for closely related incidents.
Alleged: You.com , xAI , Perplexity , OpenAI , Mistral , Microsoft , Meta , Inflection , Google と Anthropic developed an AI system deployed by TigerWeb , Storm-1516 , Russian state media , Pravda disinformation network , Portal Kombat , John Mark Dougan と Government of Russia, which harmed Truth , Policymakers , Media consumers , Journalists , Journalism , Governments , Government of Ukraine , General public of Ukraine , General public , Democracy , Chatbot users , Epistemic integrity , Journalistic integrity , Media integrity と National security and intelligence stakeholders.
関与が疑われるAIシステム: You.com Smart Assistant , xAI Grok , Perplexity Answer Engine , Mistral le Chat , Meta AI , Inflection Pi , Gemini , Copilot , ChatGPT 4o , ChatGPT と Anthropic Claude
インシデントのステータス
Risk Subdomain
A further 23 subdomains create an accessible and understandable classification of hazards and harms associated with AI
4.1. Disinformation, surveillance, and influence at scale
Risk Domain
The Domain Taxonomy of AI Risks classifies risks into seven AI risk domains: (1) Discrimination & toxicity, (2) Privacy & security, (3) Misinformation, (4) Malicious actors & misuse, (5) Human-computer interaction, (6) Socioeconomic & environmental harms, and (7) AI system safety, failures & limitations.
- Malicious Actors & Misuse
Entity
Which, if any, entity is presented as the main cause of the risk
Human
Timing
The stage in the AI lifecycle at which the risk is presented as occurring
Post-deployment
Intent
Whether the risk is presented as occurring as an expected or unexpected outcome from pursuing a goal
Intentional


