Disinfo Update 09/09/2026

Dear Disinfo Update readers, After a summer where climate change stopped feeling like a distant projection and started showing up in our daily lives, we hope you managed to recharge your batteries. The need is real: as the German regional elections recently reminded us, disinformation continues to shape political outcomes in ways we cannot afford to ignore. More than ever, we need to explore solutions and make sure enforcement against disinformation perpetrators and enablers is moving ahead. Our team is back with the latest Disinfo Update and a few bonuses. First, #Disinfo2026 is around the corner, kicking off in less than a month in Vilnius, Lithuania. We’re releasing today the final programme, including details of the opening interview with US-sanctioned Judge Kimberley Prost from the International Criminal Court and new workshops for registration very soon. Like in previous editions, the hardest choice will be choosing which session to attend, distributed across five main topics this year: #solutions, #enforcement, #AI, #investigations and #infospaces. Get prepared, and if you haven’t confirmed your ticket, please do so! We’re close to selling out, and registrations may close earlier than planned if capacity is reached before the scheduled deadline of 22 September. In parallel, we continue to provide the community with access to expertise and analysis. We’re kicking off the new season of our webinar series next week, and our policy team is releasing a new blog post unfolding a French ruling concerning sanctioned entities, disinformation, and the role of Cloudflare and online intermediaries – a case that speaks directly to the enforcement questions we have been tracking all year. Finally, we’re also resuming our community meetups in Brussels and will have more updates on our community outreach very soon! Wishing you a good read and looking forward to meeting you very soon! AI DISINFO WATCH AI and disinformation: a summer defined by regulation. New transparency rules landed, with risk-assessment requirements close behind. Since 2 August, AI providers and deployers face new obligations under Article 50 of the EU AI Act and California’s Artificial Intelligence Transparency Act (CAITA) that was quickly followed by criticism.The most contentious case: Anthropic’s plan to apply the rule to Claude-generated text via secret-key patterns, which technology blogger John Gruber called “a perversion of writing” that risks degrading prose and flagging human text as AI-made. AI companies are also required to provide publicly available detection tools that let people verify whether a piece of content was created or edited with their systems.According to an Indicator report, however, only some of them are complying. Separately, ChatGPT was designated a Very Large Online Search Engine (VLOSE) under the Digital Services Act (DSA), recognising its role as a search engine with over 45 million users and triggering stricter transparency and risk-mitigation duties. In the meantime, new events show the need for further scrutiny. OpenAI rolled out ads in 31 European countries, saying it won’t profile users by chat history. Nevertheless, privacy concerns are already mounting. On another note, RSF found most mainstream chatbots will retrieve content from sanctioned Russian state media if asked, exposing a loophole in sanctions enforcement. And what happens with extremist content? Factiverse audits found LLMs often condemn violence just long enough to pass safety filters, then validate the ideology behind it. Factiverse will expand on these findings on 17 September in our webinar, “Reject Then Launder”: Auditing AI & Disinformation in Crises. Don’t miss it! One final insight: the problem isn’t only the models but the humans behind the prompts: an analysis by investigator Henk van Ess found that chatbots echo users’ false premises back as facts. - Guidelines on Transparency of AI-Generated Content (European Commission) - How Claude’s text watermarking works (Anthropic) - Anthropic says it will watermark text generated by its AI models (TechCrunch) - Anthropic’s Watermark Text Adulteration in Claude Is a Perversion of Writing (Daring Fireball) - AI generators are now required to offer detection tools: we tested them and they need work (Indicator) - Commission designates ChatGPT under Digital Services Act (European Commission) - Artificial intelligence: Chatbots make it easier to access content from Russian media outlets sanctioned by the European Union (LeMonde/RSF) - ChatGPT Ads expands across Europe (OpenAI) - How ChatGPT becomes ad-free again, even without an expensive subscription (Der Spiegel) - Stop blaming AI chatbots. You are part of the problem (Digital Digging) Explore more in our recently updated AI Disinfo Hub, where we gather the latest research, tools, and analysis on AI-driven disinformation. INFORMATION CRISIS & ENFORCEMENT The Digital Services Act (DSA) contains the tools to respond to crises and enforce compliance; the question is whether regulators are actually using them. Tens of thousands of migrants crossed from Morocco into Ceuta on 30-31 July in an incident that left dozens dead. Allegedly, the tragedy happened due to online disinformation campaigns. Subsequent research confirmed that Russia exploited the crisis online afterwards, amplifying harms in the information and physical spaces. Despite the clear online components, it appears that the Digital Services Board, a body gathering both the European Commission and the 27 national regulators, did not call for the triggering of the DSA’s crisis mechanism, a tool that would require deeper investigation and, possibly, mitigation. Telegram shows a different version of the enforcement gap. An investigation by Democracy Reporting International (DRI) found 19 Telegram channels posing as local news, pushing Kremlin-aligned content. However, the platform is subject to lighter DSA mitigation measures given that it claims its Monthly Active Users figure is just under the 45-million-user threshold for Very Large Online Platforms (VLOPs). Could it soon be the same for X? X’s own declining EU user base raises the same question from the other direction: how close would a platform need to drift toward that threshold before its VLOP status came up for review? After meaningful transparency, meaningful enforcement? The DSA itself is now facing a different kind of scrutiny. On top of whether it’s being enforced, the question now is whether enforcement actually changes anything. The law has produced €870 million in fines against X, Temu and AliExpress, plus findings against Meta and TikTok – real money, but still far below the maximum penalty the DSA allows of up to 6% of a company’s global annual revenue. Whether harder enforcement would actually change business models built on scale and engagement remains an open question the Commission has yet to test. Talking about hitting harder: we recommend reading this portrait of Johanna Brousse, the French prosecutor leading the criminal investigation into X and wanting to bring Elon Musk in court, who has gone public with her account in a new book, Cybercombattante (“Cyberfighter”). On the sanctions side, the Council of the EU announced that its latest package against Russia designates eight individuals specifically for war propaganda. - Spanish PM blames Russia, Israel for disinformation in Ceuta migrant crisis (Reuters) - Brussels finds no evidence that the Ceuta crisis originated from disinformation by foreign powers (Demócrata) - The final text of the Digital Services Act (Article 36 of the DSA) - Telegram disinformation network raises questions over EU platform rules (Digital Watch Observatory) - An inauthentic network of Telegram channels and why the DSA can do little about it (DRI) - X continues to lose EU users (Social Media Today) - Digital Services Act fines hit €870 million: X, Temu and AliExpress (mrkt30) - Europe’s summer of Digital Services Act enforcement targets platform design (Tech Policy Press) - Elon Musk: Who is Johanna Brousse, the Frenchwoman who dares to take him on in court? (La Nouvelle Tribune) - 21st package of sanctions: EU hits Russian energy, financial services and crypto hard (Council of the EU) ON THE FRONTLINES Russia’s influence apparatus is working on two levels: shifting tactics abroad, and a stable architecture at home. Tactically, operations keep adapting. According to CeMAS, Operation Matryoshka flooded German social media ahead of Sunday’s elections in Saxony-Anhalt, Mecklenburg-Vorpommern and Berlin, falsely accusing 53 politicians across five democratic parties of crimes spanning from murder to corruption, and deploying celebrity deepfakes under hashtags like #StopMerzToday. An investigation by European External Action Service (EEAS), reported by EUvsDisinfo, traced foreign recruits lured into Russia’s army via Meta ads, despite Meta’s 2022 pledge to ban Russian advertisers. Recorded Future found that CopyCop impersonated TechCrunch to undermine a US-Armenian AI data center; Graphika and Code for Africa identified 44 fabricated journalists laundering pro-Russia narratives across African outlets; and leaked documents obtained by Organized Crime and Corruption Reporting Project (OCCRP) show that Russia’s Federal Security Service FSB is building an AI tool to merge media monitoring with internal databases. TruthAfrica reports that, in Mali, Russia recasts Africa Corps civilian-casualty reports as French and Ukrainian interference, even as Ghana, deepening ties with Moscow, raises the 62 citizens killed after being recruited into the war. Structurally, a report from NATO’s Strategic Communications Centre of Excellence finds Kremlin propaganda has become an adaptive ecosystem sustaining domestic war support. A study published in Media, War & Conflict on Rossiya 1’s Vesti nedeli shows one technique: soldiers’ deaths personalised as heroic sacrifice, civilian deaths staged in forensic detail as undeniable enemy atrocity. We tackle this kind of architecture in our 24 September webinar, How Russia Accidentally Lost Hundreds of Bombs: What Its Media Reveal About the Kremlin’s War Effort – join us! Russia isn’t the only one refining these tactics. According to French media Next.ink, Israel’s government built a fake think tank and a fake peace group operating on US soil, using direct support from French communications agency Havas, to seed content chatbots like Perplexity have already cited. Separately, Popular Information found that a Senate GOP’s dark-money arm runs six fake local news sites with ownership buried in fine print, raising questions about the banalisation of such information manipulation sets. Meanwhile, the US’s own capacity to track foreign disinformation has been gutted: MIT Technology Review and Type Investigations found the Trump administration dismantled R/FIMI, the State Department unit tasked with tracking foreign disinformation, following a “censorship-industrial complex” narrative that has also been used to justify visa bans on European DSA advocates. A different kind of contest is playing out alongside these: Le Monde reports the US seized domains a Chinese hacking group used against NASA and federal agencies, cyberespionage, not influence, but part of the same broader fight over information infrastructure. - Disinformation ahead of state elections: how the Operation Overload/Matryoshka campaign works (CeMAS) - From social media to the front line: Russia’s foreign recruitment pipeline (EUvsDisinfo) see also Meta’s 2022 commitment to ban ads from Russian advertisers - CopyCop targets AI investment in Armenia (Recorded Future) - Umbrae Ex Machina: fabricated journalists planted across African outlets to advance Russian interests (Graphika/Code for Africa) - Russian security service planned AI tool to track media (OCCRP) - Africa Between Fronts (TruthAfrica) - Crisis, Control, Crusade: Russia’s propaganda architecture (StratCom COE) - Staging (un)acceptable death: techniques in Russian war propaganda (Sage Journals) - For Israel, communications agencies create a fake think tank to influence chatbots (Next.ink) - Senate Republicans launch network of fake news sites (Popular Information) - How ideas of a vast censorship network moved from the online fringe to Trump policy (MIT Technology Review) - US seizes domain names suspected of serving Chinese espionage (Le Monde) CIVIL SOCIETY, MEDIA AND DEMOCRATIC RESILIENCE Democratic institutions and civil society groups are pushing back against the erosion of information integrity. Call for input: CheckFirst and The European Fact-Checking Standards Network (EFCSN) are running a short survey on the most pressing problems facing Europe’s information integrity community, to help shape future advocacy and priorities. Take the survey! Nineteen members of the US Congress wrote to Mark Zuckerberg on 17 August, warning that Meta’s rollback of election-integrity safeguards leaves the platform poorly prepared for the 2026 midterms. Since ending its third-party fact-checking program in January 2025, Meta has replaced it with Community Notes, while reportedly automating up to 90% of its privacy and integrity risk reviews. US fact-checkers are moving to fill part of that gap themselves: FactCheck.org, Lead Stories, PolitiFact, Factchequeado and AFP Fact Check have launched a joint newsletter tracking midterm disinformation trends. The same platform gaps show up beyond the US. Ahead of Brazil’s 4 October election, Amnesty International tested Facebook’s ad moderation with 14 mock disinformation ads: eight were approved outright, and the rest were merely flagged for identity verification, not rejected for their content. Ads calling for military intervention fared particularly well, with three of four passing through undetected. Some groups face a version of this fight that goes beyond any single election. A new paper argues gendered disinformation – the coordinated targeting of women, trans and non-binary people through deepfakes and harassment – remains poorly addressed by the DSA, despite one EU survey finding 32% of women in politics had faced violence during their careers. - 19 members of Congress write to Zuckerberg over Meta’s election-integrity rollback (letter) - US fact-checkers launch collaboration debunking deceptive 2026 midterm ads and election disinformation (Poynter/IFCN) - Amnesty International experiment exposes Meta’s failure to catch election-related disinformation ads in Brazil (Amnesty International) - Silenced by Design: gendered disinformation and its threat to democracy (CDT) BRUSSELS CORNER The European Commission announced the designation of ChatGPT (a generative AI service) as a very large online search engine (VLOSE), and Roblox (an online game development community/platform and Reddit (an online forum) as very large online platforms (VLOPs) on 31 August. This means that they will be subject to additional regulatory obligations after a four-month bedding-in period. The European Commission appears to have announced “Mission Accomplished” regarding the use of fines to ensure compliance of big tech with EU rules, according to an article published in Politico. Furthermore, only four years after the Digital Services Act (DSA) was adopted, the Commission confirmed, without hesitation, that it “has not hesitated and will not hesitate to open non-compliance proceedings whenever necessary depending on concrete circumstances”. Meanwhile, the Dutch Media Authority took a swipe (EN) at the demands for age restrictions on use of social media platforms, arguing that problems are due to the “design, functioning and revenue model” of the platforms (almost a copy/paste of the definition of a “systemic risk” under the DSA). The authority very sharply observes that “general age limit for social media cannot, by itself, remove these underlying factors”. 🧡 One thing we loved A First Amendment win for tech researchers. A federal court has temporarily blocked a Trump administration immigration policy that used visa denials, detention and deportation to target noncitizen researchers, fact-checkers and trust-and-safety workers, simply for the work they do. Chief Judge James Boasberg found the policy likely unconstitutional, ruling it discriminates against a specific viewpoint by sweeping legitimate research and advocacy into the category of “foreign censorship.” The Coalition for Independent Technology Research, represented by the Knight First Amendment Institute and Protect Democracy, called it a major victory in the fight to protect independent platform research from government intimidation. The case continues, but for now, researchers who study how tech shapes our societies can keep doing that work without the fear of losing their visas over it. 📚 Recommended pick Raquel Miguel, Senior Researcher at EU DisinfoLab, recommends the book Prophecy: Prediction, Power, and the Fight for the Future, from Ancient Oracles to AI by Carissa Véliz, a brilliant reflection on prediction throughout history: how it determines, strips individuals of agency, exerts control over them, and ultimately harms democracies. After tracing the history and philosophy of divination and forecasting, Véliz argues for developing an ethics of prediction. At this point, our readers might be wondering: what is a book like this doing in a newsletter about disinformation? As the counter-disinformation community knows, a core principle of fact-checking is that forecasts can’t be verified, since they refer to events that haven’t happened yet. But what happens when forecasts deceive or mislead, masquerading as truths? When they have the power to manipulate beliefs and behaviour, and to cause real harm? Recognising the connections between truth, prediction, and disinformation casts contemporary phenomena, such as prediction markets, in a new light, and offers a new prism to observe the greatest instrument of prediction (and surveillance) ever created: Artificial Intelligence. Because what is AI, if not an instrument of prediction through statistical patterns? What are its outputs, if not predictions? In this way, Véliz builds an inspiring ethical guide on AI based on its most fundamental root: prediction. 💡 Tip of the week A catalogue for cross-border investigations. The Organized Crime and Corruption Reporting Project (OCCRP) has built a Catalogue of Research Databases, a searchable collection of public records by region and country, useful for tracing entities, individuals and networks across borders in your next investigation. Explore it here. 🧰 Tools & learning Tracking authoritarian interference. Authoritarian Interference Tracker by Institute for Strategic Dialogue (ISD) catalogues documented interference operations by authoritarian states against democracies, across cyber, kinetic, information, financial, political and economic dimensions. Explore the tracker. Our latest work Cloudflare, sanctions, and the DSA. Our policy team has published a new blogpost on a recent French court ruling touching on sanctioned entities, disinformation, and the role of online intermediaries.This blogpost, although very long, is a very simplified summary of the legal wranglings about an order placed on Cloudflare to use its DNS service in France to block access to sanctioned Russian outlets. 🤝 Community meetups Join us this Wednesday evening, 9 September, in Brussels. Our informal community meetups bring together researchers, journalists, policymakers, technologists and civil society practitioners for an evening of exchanging ideas, making new connections, and unwinding together. If you’d like to join us, simply reply to this email and we’ll send you the location. 💬 EU DisinfoLab webinars - 17 September. “Reject then launder”: Catching AI Models validating extremist narratives in crisis - 24 September. How Russia accidentally lost hundreds of bombs: What its media reveal about the Kremlin’s war effort Missed a session? Watch our past webinars. 🗓️ Events on our radar - 17 September. The International Democracy Day Brussels Conference, by EED & Partners (Brussels, in-person) - 22 September. Disinformation Space: From Prompt to Propaganda, Generative AI and the New Disinformation Lifecycle (Dresden, hybrid) - 23 September. BBC Monitoring: From Gaza to Brussels. How the Middle East conflict is reshaping perceptions of Europe (Brussels, in-person) - 5 October. What Echoes Next: RESONANT Final Event. Enhancing resilience against Foreign Information Manipulation and Interference (FIMI) targeting diaspora communities, (Brussels, in-person/hybrid) - 5-7 October. Reporting on AI Intensive – training for journalists covering AI and technology (online) - 7–8 October. #Disinfo2026, by EU DisinfoLab and pre-conference workshops on 6 October (Vilnius, in-person) - 14–18 October. GCJT & iMEdD’s Ideas Zone European Journalist Retreat on Trauma, Resilience and Ethical Reporting, by GCJT & iMEdD. (Laconia, in-person) - 21–23 October. 1st Annual Digital Publics Conference, by DSI Community (Zurich, in-person) - 11–12 November. FACTCHECK Final Conference: Knowledge Is Power in the Age of Information Disorder (Porto, in-person) - 23–25 March 2027. Cambridge Disinformation Summit 2027 (Cambridge, in-person) 🤝 Jobs & opportunities - ISD. Open positions - NewsGuard. Staff Reporter - CCDH. Digital Fundraising and Growth Strategist - Alice (former ActiveFence). Multiple positions - Moonshot. Multiple positions - CDT. Academic Year Externship - ILGA-Europe. Advocacy Director - European Endowment for Democracy (EED). Monitoring, Evaluation & Learning (MEL) Specialist - Tech Policy Press. 2027 fellowship cohort - Thomson Reuters Foundation. Civil Society Lead - Centre for Information Resilience. Ukrainian-Language Editor (Contractor) - IAPS (Institute for AI Policy and Strategy). Spring 2027 AI Policy Fellowship, fully funded, 3-month program, Washington D.C./London/remote. - Funding opportunities - Coimisiún na Meán (Ireland) is inviting applications for its Media Literacy & Countering Disinformation Micro Funds, up to €3,000 in micro-grants for community-based media literacy projects (Strand A), or up to €25,000 for collaborative innovation projects at regional/national level (Strand B). Applications close 9 October 2026, 12:00; an info webinar for prospective applicants runs 16 September, 12:00 (register here). How to apply. - ARTE.tv and Guardian Documentaries are jointly funding a short documentary on “Democracy at Stake in Europe”, €11,500 in production funding plus editorial support, with the winning film distributed across both platforms in seven languages. Applications close 14 September. Apply here. Have something to share – an event, job opening, publication? Send your suggestions via the “get in touch” form below, and we’ll consider them for the next edition of Disinfo Update.