by WorldTribune Staff, October 21, 2026 — Real World News
A European Union-funded project called HATEDEMICS, which began on April 1, 2024, stated its primary objective was to combat online hate speech, disinformation, polarization, and conspiracy theories using advanced AI-based technologies.
Critics say HATEDEMICS not only turbocharged censorship but seeks to redirect dialogues.
The project, funded under the EU’s Citizens, Equality, Rights and Values (CERV) program, sought to:
• Monitor, detect, and record hate speech and disinformation on social media.
• Estimate “HATEDEMICS risk” by calculating exposure to potential hate and misinformation.
• Create dialogue-based, human-in-the-loop counter-narratives.
• Automatically assess behavioral changes determined by the counter-narrative intervention.
The AI platform would then provide non-governmental organizations (NGOs), fact-checkers, and public authorities across EU Member States with targeted, automated tools to deploy and measure the effectiveness of digital counter-narrative campaigns.
Ultimately, however, it was humans with an agenda who were controlling the AI’s final responses, investigative journalist Natalie Winters reported on Oct. 20.
According to project materials obtained by Winters, the EU-funded HATEDEMICS platform has already “processed more than ten million social media messages.”
Winters noted:
“Now the institution coordinating that work is developing a successor that plans to tailor counter-narratives to individual users’ emotions and perspectives.
“The two projects — HATEDEMICS and ECLIPSE — carry approximately €4.085 million in combined EU funding. Their documents describe an effort to give researchers, public authorities and organized civil-society groups increasingly sophisticated tools for intervening in online conversations.”
The implication, Winters added, “is straightforward: public funding was supporting the capacity to enter a conversation with a machine-assisted rebuttal — and a stated ambition to evaluate what happened afterward.”
Supporters of HATEDEMICS essentially confirmed this:
The European Association for Local Democracy (ALDA) reported in August:
“Most users reported that HATEDEMICS significantly reduced the time needed to identify and respond to online hate, in some cases by close to half. The majority also felt that the platform improved their ability to carry out their work effectively.
“This matters particularly for professionals who regularly navigate hostile, discriminatory or misleading online content. Saving time is not simply a question of efficiency: it can mean having more capacity for prevention, education, fact-checking and meaningful engagement with communities.
“The project also goes beyond monitoring and response. Through its platform, HATEDEMICS provides educational activities and resources designed to strengthen knowledge and skills around hate speech, disinformation and responsible digital participation.”
The work plan for ECLIPSE, the follow-up to HATEDEMICS, “calls for two rounds of pilots involving law enforcement, civil society and international organizations,” Winters wrote. “It envisages tools for analyzing perceptions and networks, generating responses, and training practitioners. Its plans also include compatibility with existing systems and expansion beyond the initial trials.”
Winters continued: “The institutional arrangement matters. Agencies with enforcement responsibilities are helping develop tools alongside organizations that produce responses to online speech. That makes the boundaries between classification, persuasion and enforcement a central accountability issue.
“Who Controls the Reply?
“ECLIPSE says its purpose is to address already unlawful content. It disclaims monitoring private communications, promises human review and emphasizes counterspeech over removal. …
“The records establish funded development, large-scale message processing and plans for personalized intervention. They leave consequential questions unanswered: how many live replies have been sent, whether recipients will be told that AI or publicly funded organizations helped produce them, and what evidence demonstrates behavioral effects.
“Before these tools expand, the public should see the pilot protocols, selection criteria and impact assessments — and know whether the apparently ordinary reply in a comment thread was shaped by a publicly funded system designed to change the conversation.”
