Search

Iran’s cognitive warfare – when AI becomes the operating layer of state propaganda Iranian state-linked influence operators were not using artificial intelligence simply to write faster social-media posts. According to Anthropic’s September 2026 threat intelligence report, three separate operations connected to Iranian propaganda institutions were using Claude to help construct the machinery behind what the operators themselves called “soft war” and “cognitive warfare” – non-military efforts intended to shape public opinion inside Iran and abroad. Anthropic linked the activity to the Islamic Culture and Communications Organization (ICCO), which operates under Iran’s Ministry of Culture and Islamic Guidance; the Islamic Propaganda Office of Khorasan Razavi; and the Bina Cultural Observatory, part of the Islamic Propaganda Organization. The company said each operation was run by someone working within or on behalf of the named institution. The attribution was based on a combination of account telemetry, institutional references disclosed during conversations, branded documents and open-source corroboration. Anthropic subsequently removed the accounts. AI as an operational headquarters The most significant feature of the operation was the role assigned to AI. Claude was used to develop campaign plans, doctrine manuals, coded project portfolios, persona systems, target databases, early-warning protocols, amplification schedules and ministerial planning documents. Anthropic described the model as the principal administrative and operational layer supporting the three campaigns, allowing relatively small groups of operators to create organisational structures and planning material that would otherwise have required a much larger staff. This distinction matters. The reported use of AI went considerably beyond content generation. Operators were using the model to help organise how influence activity should function: which audiences to target, which identities should deliver particular narratives, how official material should be repackaged and when amplification should take place. Human operators still determined objectives and supplied the underlying political doctrine; AI helped translate those decisions into an operational system. Anthropic found that all three operations explicitly connected their activity to Jihad al-Tabyin, often translated as “explanatory jihad”, a concept used within the Iranian state system to frame information activity as both a strategic and ideological responsibility. In internal planning material examined by the company, this doctrine was not merely rhetorical: it informed manuals and campaign structures used by the operators. ‘Not the narrator, but the director’ One sentence contained in the ICCO material captures the model particularly clearly. According to Anthropic, an operator described the role of Iranian cultural attachés as being “not to be the narrator, but the director” of the narrative. The distinction reveals the logic of attribution laundering. Instead of publishing an overt state message and asking foreign audiences to believe it, the operator attempts to create the appearance that the same narrative emerged independently from journalists, activists, foreign writers or ordinary citizens. Anthropic found that Claude was used specifically to make content appear to come from foreign authors or independent media organisations and to design hashtag campaigns that looked grassroots rather than centrally organised. The ICCO operation reportedly used the organisation’s international cultural network as part of this structure. Anthropic found ministerial-level documents carrying official ICCO branding, including a nine-part international influence portfolio. This makes the operation particularly significant because the infrastructure was not limited to anonymous social-media accounts; it intersected with formal state institutions and an existing network of cultural representation abroad. A multilingual content factory A second operation was associated with the Islamic Propaganda Office of Khorasan Razavi and what Anthropic described as a “cognitive warfare command room” operating from a seminary in Mashhad. Its organisers ran a multi-province content production system known internally as Manjanegh, or “Catapult”. According to Anthropic, dozens of activists were involved in repackaging publicly available reporting from Iranian security institutions under different personas so that the material no longer appeared directly connected to those institutions. The network used Claude to transform official government intelligence bulletins into customised material in Farsi, Arabic, Urdu, Malay, Spanish and English, while planning expansion into a total of 20 languages. Distribution extended across Iranian platforms such as Eitaa, Bale and Rubika and international services including X, Instagram, Telegram, TikTok and YouTube. Anthropic also documented paid amplification across more than 100 Iranian channels, including channels associated with IRGC narratives. The scale should nevertheless be interpreted carefully. Anthropic documented production infrastructure and some real-world dissemination, but it did not demonstrate that all planned campaigns were fully deployed or that they meaningfully changed public opinion. The company placed the operation in Category Three of its Breakout Scale, indicating multi-platform activity with content observed in external distribution channels, rather than evidence of strategic impact on entire populations. From security bulletins to apparently independent voices The Bina Cultural Observatory operation provides another example of the same architecture. Anthropic linked a director-level official at Bina to activity in which Claude generated messaging in the voice of an IRGC spokesperson. During the 2026 US–Israel–Iran war, the network also attributed false claims to Western institutions including CSIS, Brookings and RAND. The apparent objective was to make Iranian state-aligned claims look as though they had been independently validated by respected foreign research organisations. That technique is more sophisticated than simply publishing propaganda under a false name. It attempts to manufacture a chain of external validation: official information is transformed into apparently independent commentary, which can then be recirculated as evidence that outside observers have reached the same conclusion. Anthropic also found target databases naming international officials and Iranian opposition figures, as well as aggressive counter-narrative material directed against the Bahá’í community. These findings show that the infrastructure was designed not only for broad messaging but also for targeted influence activity against specific groups and individuals. What AI changed The Iranian operations described by Anthropic did not create a new doctrine of influence. Iran has used cultural institutions, official media, aligned organisations and covert or semi-covert online networks for years. What AI changed was the economics and organisational capacity of that model. A relatively small team could use the system to draft doctrine, create personas, translate material, structure target databases, produce campaign calendars and generate multiple versions of the same message for different audiences. The operator remained responsible for intent, political direction and selection of targets. AI reduced the amount of human labour required to turn those decisions into a functioning influence operation. The most revealing part of the case is therefore not that an Iranian propaganda network used a chatbot. It is that AI was being treated as part of the back office of information warfare – a system for converting state doctrine into apparently decentralised, multilingual and plausibly independent voices. Sources Anthropic — Detecting and countering misuse of AI: September 2026 https://www.anthropic.com/threat-intelligence-report-september-2026 Iran International — Iran state-aligned accounts used Claude to push IRGC narratives https://www.iranintl.com/en/202609105561 BNE IntelliNews — Yemen missile cell used AI to write guidance software, Anthropic says https://new.intellinews.com/articles/yemen-missile-cell-used-ai-to-write-guidance-software-anthropic-says-467296 This is Beirut — Anthropic report details Iranian propaganda operations https://thisisbeirut.com.lb/news/politics/anthropic-report-details-iranian-propaganda-operations-and-yemen-missile-development-using-claude Anthropic report — PDF mirror https://static.foxbusiness.com/foxbusiness.com/content/uploads/2026/09/anthropic-detecting-and-countering-091026-1.pdf Keywords #DISINFORMED #Episode11 #Iran #IranianInfluence #IranianPropaganda #CognitiveWarfare #SoftWar #JihadAlTabyin #Disinformation #InformationWarfare #InfluenceOperations #ForeignInterference #FIMI #ArtificialIntelligence #AI #AIPropaganda #GenerativeAI #ClaudeAI #Anthropic #SyntheticMedia #NarrativeManipulation #NarrativeLaundering #AttributionLaundering #PersonaNetworks #Astroturfing #SocialMediaManipulation #DigitalInfluence #StatePropaganda #IslamicCultureAndCommunicationsOrganization #ICCO #IslamicPropagandaOrganization #KhorasanRazavi #IRGC #MultilingualInfluence #Targeting #Amplification #OSINT #WRLD

Fake NV – how a cloned newsroom turned an AI video into a $140 million corruption scandal For two days in September, a website that looked almost indistinguishable from one of Ukraine’s best-known news organisations carried an extraordinary corruption story. It claimed that Ukraine’s National Anti-Corruption Bureau had raided former defence procurement chief Arsen Zhumadilov and discovered more than $140 million in cash packed inside refrigerator boxes. One box supposedly carried the words “For the Servant of the People” – an apparent reference to President Volodymyr Zelenskyy’s political party. The raid never happened. The photographs and video presented as evidence were fabricated, and the news organisation reporting the story was not the New Voice of Ukraine but a cloned version of it. The episode is a compact example of an increasingly effective disinformation technique. Rather than inventing a new propaganda outlet and trying to build its credibility from scratch, an operator can simply steal the identity of an established newsroom and place fabricated material inside an otherwise convincing copy of its editorial environment. A newsroom copied almost in full The counterfeit site appeared at nvukraine.com, while the authentic New Voice of Ukraine operates at nv.ua and english.nv.ua. Registration records examined by fact-checkers show that the false domain was created on 21 September 2026, two days before the fabricated article appeared. According to NV’s own investigation, the imitation went much further than copying a logo: the operators reproduced the site’s visual design, sections, branding, contact information and genuine editorial material. A snapshot of legitimate NV content from 23 September appears to have been copied onto the counterfeit site, into which the operators inserted a single false article. That article claimed that NABU had searched premises connected to Zhumadilov, who had headed Ukraine’s Defence Procurement Agency until August, and seized $140 million in cash. The story was published in several languages, including Ukrainian and English, suggesting that the operation was intended not only for Ukraine’s domestic information space but also for foreign audiences following the war and Western support for Kyiv. AI supplied the supposed evidence The false article was supported by a video purportedly showing the money discovered during the raid. NABU said no such search or procedural action had taken place and identified the footage as fabricated using artificial intelligence. Independent fact-checker Lead Stories found no credible Ukrainian or international reporting confirming either the raid or the alleged seizure, while searches of NABU’s own published material produced no evidence supporting the claim. The operation therefore combined two separate forms of synthetic credibility. The AI-generated footage supplied what looked like visual proof, while the cloned NV website supplied institutional proof: an apparently reputable Ukrainian newsroom supposedly confirming that the event had happened. Neither element needed to withstand close scrutiny for long. Together, they only needed to appear credible enough to be copied, quoted and redistributed before the deception was exposed. From a fake Ukrainian source into the Russian information ecosystem On 24 September, the story began moving through large pro-Russian Telegram channels. Ukraine’s Center for Strategic Communications traced early amplification to Golos Mordora, with roughly 149,000 subscribers, and Militarist, with around 277,000. Within less than an hour, other channels including VOBLA, Rossiya Seychas and Ostashko! Vazhnoye were circulating versions of the same claim. It subsequently appeared on X and on Russian-language websites including e-news.su and pressa24.ru. Babel calculated that posts carrying the fabrication had accumulated approximately 370,000 views by the morning of 25 September. The story also moved beyond the Russian-language ecosystem. English-language versions appeared in online communities including Lemmy and Hexbear, while accounts on X shared direct links to the counterfeit NV article. NV later reported that larger Russian media and propaganda outlets, including RIA Novosti, Life, Vesti, Ukraina.ru and Mail.ru, had reproduced or discussed the claim. Kirill Dmitriev, head of the Russian Direct Investment Fund and a Kremlin representative involved in contacts with the United States, also shared it on 25 September. This amplification is significant, but it should not be confused with proof of authorship. Publicly available evidence shows how the false story entered and spread through a Russian and pro-Russian information ecosystem. It does not establish which organisation registered the cloned domain, generated the video or directed the operation from the outset. Why copy NV? The operation relied on source laundering. A claim originating directly from a Russian Telegram channel that Ukraine’s defence procurement system had concealed $140 million would immediately carry an obvious political context. A claim apparently uncovered by an established Ukrainian newsroom looks very different. Once the fabricated NV article existed, subsequent accounts could cite what appeared to be an independent Ukrainian source rather than present the allegation as their own. Each repost moved the story another step away from its manufactured origin. The use of an English-language version strengthened that mechanism for audiences outside Ukraine. The underlying narrative was also carefully chosen. Corruption in Ukrainian defence procurement is a genuine subject of public scrutiny and investigative reporting, so the fabricated story did not require audiences to accept an entirely unfamiliar premise. It inserted a false event into a real area of political concern. That mixture of authentic context and fabricated evidence can be more persuasive than a completely invented story because much of the surrounding information feels recognisable. What can – and cannot – be attributed Ukrainian authorities and NV describe the episode as a Russian information operation, and its distribution can be documented extensively across pro-Russian and Russian media channels. The available public evidence, however, does not identify the organisation that created the counterfeit website, nor does it link the operation forensically to a named Russian intelligence service, military unit or state contractor. The most precise description is therefore a Russian-aligned disinformation operation whose fabricated material was systematically amplified through the Russian information ecosystem. That distinction matters. Distribution can be observed directly; command-and-control is harder to establish and should not be inferred without evidence. The newsroom itself becomes the deepfake The most important feature of the operation was not the $140 million figure or even the AI-generated video. It was the counterfeit newsroom surrounding them. The operators copied a real publication, populated it largely with genuine material and inserted a single false story, allowing the fabrication to borrow the accumulated credibility of journalists who had nothing to do with it. This points to a broader evolution in synthetic media. AI can fabricate an image, a voice or a video, but an influence operation can now fabricate something larger: the source that supposedly verified the evidence. In the Fake NV operation, the article, the video and the website reinforced one another, creating not simply a false story but a manufactured chain of credibility. Sources The New Voice of Ukraine — investigation into the cloned website https://english.nv.ua/nation/nv-discovers-fake-copy-of-website-with-planted-story-cites-russian-links-50644601.html The New Voice of Ukraine — Russian amplification of the $140 million fabrication https://english.nv.ua/russia-spreads-fake-140-million-ukraine-corruption-story-during-dmitriev-u-s-talks-50644647.html Lead Stories — independent fact-check https://leadstories.com/hoax-alert/2026/09/fact-check-fake-story-claims-140-million-found-at-home-of-ukraines-zhumadilov-marked-servant-of-the-people-nv-website-originated-it.html Babel — AI video and distribution analysis https://babel.ua/en/news/130469-a-video-is-sharing-online-about-140-million-found-in-the-possession-of-the-ex-head-of-the-defense-procurement-agency-this-is-an-ai-video-distributed-by-a-clone-of-ukrainian-media Center for Strategic Communications and Information Security — initial debunk https://spravdi.org/en/ai-fake-about-140-million-amid-us-talks-how-russians-impose-the-image-of-corrupt-ukraine-on-the-west/ Center for Strategic Communications and Information Security — NABU confirmation https://spravdi.org/u-merezhi-poshyryuyut-shi-fejk-pro-nibyto-obshuky-nabu-v-eksgolovy-agencziyi-oboronnyh-zakupivel/ Detector Media — distribution and NABU response https://detector.media/infospace/article/252806/2026-09-25-u-merezhi-pidrobyly-sayt-nv-dlya-poshyrennya-feyku-pro-obshuky-v-eksochilnyka-agentsii-oboronnykh-zakupivel/ Keywords #DISINFORMED #Episode10 #FakeNV #NewVoiceOfUkraine #Ukraine #Russia #RussianDisinformation #RussianInfluence #Disinformation #InformationWarfare #InfluenceOperations #ForeignInterference #FIMI #MediaImpersonation #ClonedMedia #FakeMedia #FakeNews #ArtificialIntelligence #AI #AIGeneratedContent #SyntheticMedia #AIVideo #SourceLaundering #NarrativeLaundering #CorruptionNarrative #UkraineCorruption #DefenceProcurement #NABU #ArsenZhumadilov #Telegram #SocialMediaManipulation #DigitalInfluence #Propaganda #OSINT #MediaManipulation #WRLD

UAE / Sudan – when an influence operation tried to enter the United Nations A network of roughly 300 apparently independent social-media influencers. A human-rights organisation borrowing the identity of a real NGO. Personal dossiers on European politicians and journalists. And testimony prepared for delivery at the United Nations without revealing the state interest behind it. These were elements of an influence operation uncovered by Anthropic and disclosed in September 2026. The company says it linked the activity with high confidence to UAE government officials. Tracked as GTG-84002, the operation targeted several overlapping issues: the Muslim Brotherhood, perceptions of the war in Sudan and international mechanisms examining the United Arab Emirates' role in the conflict. What distinguishes the case is not simply its use of artificial intelligence. It is the attempt to make state-aligned messaging appear to originate from independent influencers, human-rights organisations and potentially witnesses addressing an international institution. An influence network built around 300 fake voices Anthropic identified a single actor using Claude to support a sustained influence campaign. At its centre was an AI persona called “Deadshot”, running on the operator's own platform. A master doctrine file repeatedly instructed the system to support what it described as a coordinated international effort to dismantle the Muslim Brotherhood. The operator simultaneously managed several components of the campaign, including approximately 300 inauthentic influencer accounts across social-media platforms. Internal material examined by Anthropic described the apparent independence of this network as its greatest strategic advantage. That phrase captures the central method. The objective was not simply to broadcast a political position but to conceal where that position originated. One visible example appeared on 4 June 2026, when accounts participated in a coordinated campaign using #SudanIslamists and near-identical graphics connecting Sudanese Islamists and the Muslim Brotherhood with regional instability. Anthropic says the amplification network was centrally funded and coordinated. Its investigation also found that the operator financing the social-media network was connected to the wider influence operation. A human-rights organisation that was not what it appeared to be Social media was only one layer. The operator also created a front NGO that copied the identity of a genuine Swiss organisation and used that borrowed legitimacy to publish human-rights material produced for the campaign. Anthropic's description elsewhere also refers to the identity of a real Sudanese human-rights organisation being used in connection with the preparation of testimony. The public report does not fully clarify whether these references describe the same front or separate elements of the operation, so they should not be treated as definitively identical. The underlying technique, however, is clear: political material was designed to appear as if it originated from independent civil society rather than from actors linked to a government. This is particularly significant in the human-rights environment, where the perceived independence of an organisation or witness can determine how seriously evidence is received by journalists, diplomats and international institutions. The attempt to reach the UN The most consequential part of the operation concerned the 62nd session of the UN Human Rights Council. According to Anthropic, the operator used Claude to ghost-write complete testimony intended to be delivered by two individuals during the session. The texts were prepared under explicit constraints ensuring that neither statement would mention the UAE. This did not amount to a conventional government submission. The intention, according to Anthropic's reconstruction, was for material serving the interests of a party connected to the Sudan conflict to reach an international forum through apparently independent voices. But an important limitation remains: Anthropic could not confirm that the testimony was ultimately delivered. The company similarly could not establish whether other dossiers produced by the operation reached their intended recipients or influenced policy. It therefore assessed the campaign as Category Three on the Brookings Breakout Scale – activity distributed across several platforms, but without evidence of broad public impact sufficient for a higher classification. The distinction matters. The evidence demonstrates a sophisticated attempt to influence international debate; it does not demonstrate that the attempt succeeded. Politicians, journalists and UN investigators became targets The operation was not limited to generating public content. Anthropic found that the operator researched and compiled detailed profiles of 18 members of the European Parliament and prominent journalists. Some material was prepared for direct delivery to senior UAE officials. The network also assembled what Anthropic describes as “counter-accountability dossiers” on UN Special Rapporteurs who had criticised the UAE's conduct in relation to Sudan. This places the campaign in a broader context. Since Sudan's civil war began in April 2023, the role of external powers has become an increasingly important part of international scrutiny. Sudan has accused the UAE of supporting the Rapid Support Forces, allegations Abu Dhabi has repeatedly denied. In 2025, Sudan brought a case against the UAE before the International Court of Justice, accusing it of complicity in genocide against the Masalit through alleged support for the RSF. The UAE rejected the allegations. The ICJ subsequently removed the case from its list after finding that it lacked jurisdiction because of the UAE's reservation to the relevant provision of the Genocide Convention. The ruling therefore did not determine whether Sudan's substantive allegations were true or false. That contested international environment helps explain why narratives about Sudan, human rights and accountability were valuable targets for an influence operation. AI as an operating system for influence The role played by Claude is also important to understand accurately. Anthropic did not find that artificial intelligence independently conceived or directed the campaign. Human operators established the objectives, political doctrine and targets. AI instead helped operationalise them. Across hundreds of sessions, Claude was used to transform predetermined political objectives into social-media narratives, human-rights reports, testimony, intelligence-style briefs and detailed profiles of individuals. The same system could support narrative production, target research and preparation of material for different audiences. That represents a more significant development than simply using generative AI to produce propaganda faster. Traditional influence operations require different teams to research targets, write content, manage personas, adapt messages and prepare briefing material. Generative AI can compress several of those functions into a single operational workflow. The result is not necessarily more convincing propaganda. It is potentially cheaper, faster and more scalable influence infrastructure. From fake influencers to institutional influence Anthropic says it linked the operation to UAE government officials with high confidence. Its assessment was based partly on internal material naming senior Emirati officials as intended recipients of the work, alongside other evidence available to its investigators. That remains an attribution by Anthropic rather than a judicial finding or publicly released forensic attribution by a government agency. The distinction should be retained. The larger significance of GTG-84002 lies in the architecture of the campaign. A network of fake influencers could create the appearance of public opinion. A cloned human-rights organisation could provide institutional credibility. AI-generated reports could give political narratives the appearance of research. Profiles of journalists and politicians could support more precise targeting. And ghost-written testimony could potentially carry the same message into an international institution without revealing the political interest behind it. This is a different model from the familiar troll farm. The objective is no longer simply to make propaganda look popular. It is to make state-aligned messaging look independent – and, if possible, to move it from social media into the institutions where international policy and accountability are debated. SOURCES: Anthropic — Countering misuse of AI: September 2026 https://www.anthropic.com/threat-intelligence-report-september-2026 International Court of Justice — Sudan v. United Arab Emirates https://www.icj-cij.org/case/197 International Court of Justice — press releases and case documents https://www.icj-cij.org/case/197/press-releases United Nations Geneva — Sudan v UAE proceedings at the ICJ https://www.ungeneva.org/en/news-media/news/2025/04/105241/world-court-begins-hearing-sudans-complicity-genocide-case-against UAE Ministry of Foreign Affairs — UAE response to Sudan's allegations https://www.mofa.gov.ae/en/MediaHub/News/2025/4/10/10-4-2025-UAE-UAE UN Digital Library — Sudan's letter concerning alleged UAE support for the RSF https://digitallibrary.un.org/record/4091008?ln=en UN Digital Library — UAE response concerning allegations of support for the RSF https://digitallibrary.un.org/record/4046224?ln=en Ultra Sudan — reporting on Anthropic's UAE-linked influence-operation findings https://ultrasudan.usawtiq.com/أنثروبيك-أحبطنا-عملية-تأثير-إماراتية-استخدمت-الذكاء-الاصطناعي-لاستهداف-السودان/عامر-صالح/أخبار KEYWORDS #DISINFORMED #Episode8 #UAE #Sudan #UnitedArabEmirates #UnitedNations #UNHumanRightsCouncil #HumanRights #Disinformation #InformationWarfare #InfluenceOperations #ForeignInterference #FIMI #CoordinatedInauthenticBehaviour #CIB #FakeInfluencers #FakeNGO #NGOImpersonation #Astroturfing #ArtificialIntelligence #AI #AIPropaganda #AIGeneratedContent #SyntheticMedia #SocialMediaManipulation #DigitalInfluence #Propaganda #NarrativeManipulation #InstitutionalInfluence #SudanConflict #SudanWar #MuslimBrotherhood #InternationalRelations #Geopolitics #Anthropic #ClaudeAI #OSINT #WRLD

Russia’s fake African newsrooms – propaganda designed to look local A Facebook user in Nairobi, Accra, Johannesburg or Luanda encounters what appears to be a local news page. It publishes stories about African politics, Western influence and relations with Russia. The language is tailored to an African audience and the page presents itself not as a foreign broadcaster, but as part of the local information landscape. Behind some of these outlets, however, Meta found operators based in Russia. In its 2026 threat research, the company disclosed a Coordinated Inauthentic Behaviour network targeting audiences in Angola, Ghana, Kenya and South Africa. Meta removed 37 Facebook accounts and 29 Pages connected to the operation. Around 30,000 users followed at least one of the Pages, while the operators spent approximately $7,000 on Facebook and Instagram advertising, mostly in euros and US dollars. The numbers are relatively modest. The method is considerably more important. Rather than openly distributing Russian state messaging, the network posed as local African news sources and grassroots organisations, promoting narratives about Western colonialism and political interference while presenting Russia as a credible economic and political alternative. It was an old geopolitical message delivered through a much more effective identity: not Moscow speaking to Africa, but Africans apparently speaking to one another. A news outlet that was actually an influence operation Meta's investigation found that the network attempted to conceal its Russian origin while creating media brands that appeared indigenous to the countries they targeted. Its content amplified historical grievances against former colonial powers, criticised Western involvement in Africa and promoted closer relations with Russia. One example helps illustrate how the infrastructure was assembled. A Facebook Page targeting Kenya was called Kenya Watchtower. According to OpenAI's subsequent investigation, Facebook transparency records showed that the Page had previously been named “Farmtown5”. It appears to have been acquired or repurposed rather than built from scratch as a Kenyan news organisation. Some Pages also exposed administrator locations in Russia and Ukraine. Other elements of the network had previously targeted audiences in Zambia and Namibia. This use of apparently local identities is particularly important in the African information environment. A story published by RT or another identifiable Russian outlet arrives with an obvious geopolitical provenance. The same narrative presented by something resembling an independent Kenyan, Ghanaian or South African newsroom carries a different kind of credibility. The source appears closer to the reader, and the geopolitical interest behind the message becomes less visible. Then investigators found Dr Manuel Godsin The Facebook operation was only one part of a broader information ecosystem. OpenAI investigated related activity after receiving information from Meta and subsequently banned a ChatGPT account it assessed as likely originating in Russia. OpenAI called the campaign Operation No Bell. The account was being used to generate long-form articles and social-media posts about African geopolitics. Prompts were primarily written in English, but OpenAI also observed Russian-language instructions that the user described as coming from a manager. The operator sometimes explicitly asked the model to make the material appear less AI-generated – including requests to avoid stylistic features associated with machine-generated text and to write more like a human journalist. Many articles appeared under the name “Dr Manuel Godsin”, presented as an academic with a PhD from the University of Bergen and an affiliation with an organisation called the International Centre for Political and Strategic Studies. OpenAI could find no credible evidence that Godsin existed. Searches of Norway's National Research Information Repository and the University of Bergen library produced no record of him. Investigators subsequently discovered that a photograph used to represent Godsin had appeared years earlier on a Russian professional networking site and apparently belonged to a law student in St Petersburg. The fabricated academic identity was nevertheless remarkably productive. OpenAI identified 53 online articles carrying the Godsin byline. The fiction had moved beyond fake social-media pages and into the real media ecosystem. When propaganda enters genuine newsrooms This is the most consequential aspect of the operation. The articles were not confined to websites controlled by the influence network. Some were published by genuine African news organisations, including established South African outlets. The content mixed local political issues with broader geopolitical narratives. Some pieces criticised the United States and Britain or defended Russia's role in Africa. One accused the British NGO Crisis Action of fomenting protests in South Africa. Another praised Russia's presence in the Central African Republic. Other material addressed Kenya, Angola and relations between African governments and Washington. The mechanism represents a significant evolution from the classic troll-farm model. Instead of building an audience entirely on its own platforms, an influence operator can manufacture an apparently credible expert, generate articles in his name and persuade genuine news organisations to publish them. Once this happens, the propaganda acquires something a fake Facebook Page cannot provide: the institutional credibility of a real newsroom. OpenAI assessed No Bell's social-media impact as limited. One Facebook Page had around 3,000 followers before Meta removed it, while several others had very small audiences. But the operation was more successful in placing material in established media. In OpenAI's impact framework, it approached the level at which an influence operation breaks into mainstream media. For information operators, a single article published by a recognised outlet can potentially be more valuable than thousands of impressions generated by an obviously artificial account. A much larger network of ghost journalists Subsequent research suggests that Manuel Godsin was not an isolated experiment. In August 2026, Graphika, working with Code for Africa and with support from Meta, published a much broader investigation into Russian ghostwriting operations across African media. Researchers identified 44 ghost writers and fake experts, 38 of them assessed as high-confidence fabricated personas, operating between 2021 and 2026. Their material appeared or was quoted across 138 websites and was subsequently republished through at least 113 Facebook accounts and Pages, including authentic users, coordinated inauthentic networks and official Russian communication channels. The narratives were strikingly consistent. They included criticism of France, Ukraine and the United States; attacks on organisations such as ECOWAS and the International Criminal Court; positive portrayals of Russian involvement in Africa; praise for Russian paramilitary forces; and support for sovereignty-oriented political projects such as the Alliance of Sahel States. The operation therefore extends beyond creating fake news websites. It attempts to insert Russian-aligned narratives into the legitimate African media ecosystem using identities that appear African, independent or academically authoritative. This also explains why measuring reach through the original Facebook network alone is misleading. A fabricated article may begin with an influence operator, appear in a genuine African publication, be republished elsewhere and eventually circulate without any visible connection to Russia. At that point, the provenance of the narrative has effectively been laundered. An old Russian strategy with increasingly local faces Russia's use of local intermediaries in Africa is not new. Meta documented Russian networks using African nationals as early as 2019. In 2020 it dismantled another operation involving people in Ghana and Nigeria working on behalf of individuals in Russia, with links to previous activity associated with the Internet Research Agency. More recent Meta investigations suggest that this model has continued to evolve. Networks have increasingly relied on local freelancers, social-media managers and authentic media outlets rather than exclusively operating armies of obviously fake Russian-controlled accounts. This decentralisation offers several advantages. Local operators understand language, political sensitivities and cultural references. Authentic accounts are harder to identify than newly created synthetic personas. Local media brands can also deliver narratives without immediately triggering the scepticism associated with Russian state outlets. There is no evidence that every African journalist, freelancer or publication carrying such material knowingly participates in Russian influence activity. Meta explicitly notes in its investigations that some local contractors may not know who ultimately commissioned their work. That distinction is essential. The operation's effectiveness partly depends on precisely this ambiguity between deliberate participation, commercial content placement and unwitting amplification. The objective is to make Russian narratives look African The significance of this network is therefore not its 30,000 Facebook followers or its $7,000 advertising budget. Those figures are small compared with the audiences of major African media organisations. What matters is the distribution model. Traditional foreign propaganda asks an audience to trust a foreign source. These operations attempt to remove the foreign source from the equation altogether. A Russian geopolitical narrative can be generated with AI, attributed to an expert who does not exist, published by a media organisation that does, amplified through a Facebook Page presenting itself as local, and then rediscovered elsewhere as apparently independent African analysis. By the time the reader encounters it, Moscow may have disappeared completely from the chain of attribution. The resulting information operation is therefore less about making Russian propaganda more persuasive than about making it look as though it is no longer Russian propaganda at all. In an increasingly fragmented African media environment, that may be the more important evolution to watch. SOURCES: Meta Threat Research https://threatresearch-team.github.io/indicators/meta-h1-2026-russia-based-cib-network-1/ OpenAI — Operation “No Bell” https://openai.com/index/disrupting-malicious-uses-of-ai-no-bell/ Graphika — Umbrae Ex Machina https://www.graphika.com/reports/umbrae-ex-machina News24 https://www.news24.com/southafrica/news/ai-expert-exposed-fake-kremlin-linked-analyst-planted-stories-in-african-media-20260318-0560 Vanguard https://www.vanguardngr.com/2026/03/ai-expert-exposed-fake-kremlin-linked-analyst-planted-stories-in-african-media/ Meta — Russian Coordinated Inauthentic Behaviour, 2019 https://about.fb.com/news/2019/10/removing-more-coordinated-inauthentic-behavior-from-russia/ Meta — Russian Coordinated Inauthentic Behaviour, 2020 https://about.fb.com/news/2020/03/removing-coordinated-inauthentic-behavior-from-russia/ Meta Threat Research — Russian Use of Authentic Operators in Sub-Saharan Africa https://threatresearch-team.github.io/indicators/meta-h2-2025-russia-based-cib-network-2/ KEYWORDS: #DISINFORMED #Episode7 #Russia #Africa #RussianDisinformation #RussianInfluence #Disinformation #InformationWarfare #InfluenceOperations #ForeignInterference #FIMI #CoordinatedInauthenticBehaviour #CIB #FakeNews #FakeMedia #FakeNewsrooms #MediaImpersonation #GrassrootsManipulation #Astroturfing #ArtificialIntelligence #AI #AIPropaganda #SyntheticMedia #FakeInfluencers #SocialMediaManipulation #Facebook #Meta #OperationNoBell #AfricanMedia #MediaManipulation #Propaganda #DigitalInfluence #Angola #Ghana #Kenya #SouthAfrica #OSINT #WRLD

L'Abandon (de Vincent Garenq), le film de qui raconte les 11 derniers jours de Samuel Paty, est une oeuvre majeure et implacable. Le film revient en détail sur l'émergence d'une rumeur et son amplification, sur les intimidations qui l'accompagnent et la logique mortifère des https://t.co/dhHQlN3o6t #LAbandon #SamuelPaty #cinéma #Aucuncodedepaysspécifiquen'estmentionnédansletweet.

Mali – a war where some of the people fighting do not exist In Mali's war, an apparently simple question is becoming increasingly difficult to answer: who exactly is speaking to us? A real person? A rebel organisation? A Russian influence operation? A supporter of the military junta? Or somebody who has never existed at all? In September 2026, the Dakar-based Timbuktu Institute published Des algorithmes en guerre, a report examining the use of generative artificial intelligence by actors involved in the Malian conflict. Its researchers describe the country as a potential laboratory for AI-assisted information warfare. Separatist networks are deploying synthetic fighters on TikTok. Jihadist propagandists are experimenting with AI-generated imagery. Pro-government accounts are producing large quantities of content celebrating the Malian armed forces. Russia, meanwhile, operates the most developed influence infrastructure of all: pseudo-media outlets, inauthentic accounts, AI-generated text, images and video, and systems designed to artificially amplify their visibility. Artificial intelligence did not create Mali's conflict. The war has been under way since 2012, and its participants have always fought with propaganda as well as weapons. What AI has changed is the economics of that struggle. A new “witness”, commentator, fighter, poster or video can now be produced in minutes. For audiences watching from a phone, distinguishing between genuine documentation, propaganda and material created entirely by an algorithm is becoming progressively harder. Azzghim – the fighter who cannot be killed or interrogated The most striking case documented by the Timbuktu Institute involves accounts associated with supporters of the Azawad Liberation Front, or FLA, a Tuareg separatist movement fighting the authorities in Bamako. A figure called “Azzghim” appeared on TikTok. He looks like a Tuareg fighter and regularly features in videos attacking Mali's government and Russia's Africa Corps, commenting on the conflict and promoting the cause of Azawad. There is one problem: Azzghim does not exist. Analysis of the videos identified anomalies characteristic of generative AI, including unnatural body movements, visual inconsistencies, irregular colouring and problems with audio synchronisation. According to the Timbuktu Institute, Azzghim is a synthetic character being used as the digital face of a political narrative. The propaganda advantages are considerable. An organisation no longer needs to expose a real spokesman. A synthetic personality can be designed to look exactly as its creators require, speak the appropriate language and appeal to a particular audience. Different characters could potentially be created for younger viewers, traditional communities or different ethnic and social groups. There is another advantage: a synthetic spokesman cannot be arrested, identified or interrogated. Researchers also identified accounts amplifying the material. At the time of the study, TikTok account @tawri.n.azawad had around 207,000 followers and was almost entirely devoted to publishing or repackaging content involving Azzghim. Another account, @asnan831, used a similar synthetic teenage character. An important distinction is necessary. Researchers describe these profiles as belonging to supporters or individuals associated with the FLA. The available evidence does not establish that every account is part of an operation centrally directed by the Front's leadership. The information effect, however, remains significant: a digital community can be constructed around the Azawad cause in which some of the apparent participants may not be real people at all. The jihadists are learning too The picture is different for Jama'at Nusrat al-Islam wal-Muslimin, or JNIM, the al-Qaeda-affiliated coalition that has become one of the most powerful armed groups in the Sahel. Its confirmed use of generative AI remains considerably less sophisticated. In June 2026, JNIM's Az-Zallaqa media apparatus published AI-generated posters promoting material about previous attacks. The images showed fighters in combat but contained familiar generative artefacts: buildings and vegetation merging unnaturally, blurred details and unnaturally smooth surfaces. More important than JNIM's current capabilities is what AI could add to an already sophisticated propaganda apparatus. The organisation already tailors communications to different audiences. UN reporting has documented its use of media channels for broader propaganda while other outlets increasingly distribute material in local languages, including Bambara. Generative AI could dramatically reduce the cost of this strategy, translating material into Fulfulde, Hausa, Zarma or Tamasheq, generating synthetic dubbing and producing multiple versions of the same message. The Timbuktu Institute also considers the future possibility of AI chatbots being used to personalise recruitment. That should not be presented as a capability JNIM has already demonstrated. At present, it is a risk scenario rather than a confirmed operation. Russia has the most sophisticated machine At the other end of the spectrum lies Russia's influence ecosystem. Here, AI is not an isolated experiment but another component of an infrastructure previously developed by Wagner-linked networks, Russian media operations and influence organisations operating across Africa. After Wagner's withdrawal from Mali in 2025, its military role was taken over by the Russian state-controlled Africa Corps. The information infrastructure evolved alongside it. Of particular importance is African Initiative, which presents itself as a news agency but has been identified by France's VIGINUM, the UK's Foreign, Commonwealth & Development Office and the European External Action Service as an important component of Russia's newer influence architecture in Africa. One of the most interesting elements of this ecosystem is a network researchers call AI-Freak. The operation uses generative AI to produce text, images and videos, places them on fake or apparently independent news websites and distributes them through inauthentic accounts. This is supplemented by so-called Black Hat SEO – techniques intended to artificially increase the visibility of content in search engines. Elements of this infrastructure have previously been identified by Meta and OpenAI. In 2024, OpenAI disrupted Russian accounts using ChatGPT to generate articles and comments in English, French and Russian. Some of the French-language content targeted audiences in West Africa. In a later iteration of the operation, AI models were also used to draft scripts for short videos praising the Russian Africa Corps, translate them into French and generate descriptions optimised for social media. This illustrates an important difference between the Russian approach and Azzghim. The separatist ecosystem uses AI primarily to manufacture convincing digital personalities. The Russian model increasingly resembles an industrial production line: text, image, video, pseudo-news website, distribution account, search optimisation and additional channels amplifying the narrative. The objective is not merely to convince someone of a single claim. It is to occupy as much of the information environment as possible with narratives advantageous to Moscow: Russia as an effective security partner, the West as a neo-colonial aggressor, France as a source of instability and Russian forces as defenders of African sovereignty. Yet VIGINUM also warns against equating technological sophistication with actual impact. Some Russian influence channels still attract relatively small audiences. A sophisticated operation is not automatically a successful one. The junta is building its own digital reality Mali's military authorities operate within the same contested information environment. Since breaking with France and moving closer to Moscow, Bamako has increasingly restricted independent media space while promoting a narrative built around restored sovereignty, military success and the benefits of its Russian partnership. The Timbuktu Institute highlights the TikTok account @6tm_officiel. At the time of the study, it had more than 436,000 followers and published substantial amounts of AI-generated material presenting Mali's armed forces, the FAMa, and the Africa Corps in a favourable light. One video concerning the battle for Kidal accumulated around 4.7 million views. Again, attribution requires care. Researchers describe the account as being operated by an individual based in Bamako; they do not provide evidence that it is officially managed by the Malian government. It is better understood as part of a wider pro-government information ecosystem in which the boundaries between state communications, supporters and co-ordinated propaganda can be difficult to establish. The Africa Center for Strategic Studies has previously identified the military regimes in Mali and Burkina Faso as important sources of disinformation campaigns in West Africa. Recurring narratives target France, the United Nations, ECOWAS, human-rights organisations and independent media, while presenting criticism of the authorities as part of a foreign conspiracy. Generative AI allows such narratives to be produced faster, more cheaply and in far more variations. Everyone is fighting over a different version of the same war What makes Mali particularly revealing is that multiple actors are using similar technology for very different objectives. FLA-linked networks need digital faces capable of humanising the Azawad cause and attacking Russia's presence. JNIM needs propaganda material and increasingly localised communications. The pro-government ecosystem seeks to portray Mali's armed forces as victorious and reinforce the junta's legitimacy. Russia possesses the broadest infrastructure, in which AI is merely one component of a larger system involving media outlets, websites, local partners, social-media profiles and technical amplification. The result is a war taking place simultaneously in two realities. In the first, soldiers and civilians are killed, towns change hands and the army, separatists and jihadists fight for territory. In the second, every side is trying to determine what the first reality will look like on a smartphone screen. Generative AI does not need to create a perfect deepfake to be effective. It merely needs to increase the number of competing versions of events until an ordinary viewer can no longer easily determine which one deserves to be trusted. That is why Mali matters. It is not showing us the future of information warfare. It is showing us its present. One TikTok video may feature a real soldier. Another may feature a propagandist. A third may feature a fighter who has never existed. And all three can tell completely different stories about the same war. Sources Timbuktu Institute – Des algorithmes en guerre: Comment l'IA générative rebat les cartes du conflit malien https://timbuktu-institute.org/index.php/toutes-l-actualites/item/1731-rapport-des-algorithmes-en-guerre-comment-l-ia-generative-rebat-les-cartes-du-conflit-malien AfricaNews/AFP – “Mali's information war is increasingly being shaped by AI” https://www.africanews.com/2026/09/09/malis-information-war-is-increasingly-being-shaped-by-ai/ OpenAI – research on Russia-origin influence activity and Operation Stop News https://openai.com/index/disrupting-malicious-uses-of-ai-stop-news-2024/ VIGINUM / FCDO / EEAS – technical report on African Initiative and the AI-Freak network https://www.sgdsn.gouv.fr/files/files/Publications/20250612_TLP-CLEAR_VIGINUM_FCDO_EEAS_Technical_Report_African_Initiative_EN.pdf Africa Center for Strategic Studies – Mapping a Surge of Disinformation in Africa https://africacenter.org/spotlight/mapping-a-surge-of-disinformation-in-africa/ Timbuktu Institute – analysis of AI use by JNIM https://timbuktu-institute.org/index.php/toutes-l-actualites/item/1735-artificial-intelligence-a-new-propaganda-lever-for-jnim-in-mali?print=1&tmpl=component #Disinformation #Mali #Sahel #Africa #ArtificialIntelligence #AI #GenerativeAI #AIPropaganda #InformationWarfare #InfluenceOperations #Propaganda #Deepfakes #SyntheticMedia #Azawad #FLA #JNIM #AfricaCorps #Russia #RussianInfluence #TikTok #DigitalWarfare #MediaManipulation #WRLD

Le 22 août, la température moyenne de la surface des océans a atteint un nouveau record de 21,1°C, alertant sur l'amplification du réchauffement climatique et l'augmentation des risques d'événements météorologiques extrêmes, exacerbés par le phénomène El Niño en cours. #RéchauffementClimatique #TempératureOcéanique #ÉvénementsMétéorologiques

Le documentaire Brigitte Macron, l'ombre de la rumeur, diffusé sur France 5, déconstruit la théorie complotiste selon laquelle la Première dame serait née homme, explorant son origine, son amplification sur les réseaux sociaux et illustrant la défiance croissante des citoyens envers les élites politiques. #BrigitteMacron #ThéorieDuComplot #Élites

La Norvège a enregistré son printemps le plus chaud depuis 1901, avec des températures supérieures de 2,1°C à la normale, particulièrement dans le nord, un phénomène attribué aux émissions de combustibles fossiles et à l'amplification de l'Arctique. #Climat #Norvège #Réchauffement climatique