Welcome to Disinfo Bulletin, your daily update from the EDMO network. Today's most relevant stories: 

🗳️ 🇩🇪 Disinformation about the electoral process persists in Germany. A recent false claim in the country suggests that the vote count for the far-right party Alternative for Germany (AfD) will follow stricter rules and will be invalidated if the mark on the ballot crosses a designated circle on the ballot paper. In fact, according to the Federal Returning Office, such restrictions do not exist and the only condition for a valid vote is that the mark on the ballot clearly indicates the will of the voter.

This unproven claim has been circulating for years and it is consistent with false stories recently identified in the country questioning the fairness of the EU election process, already reported in previous issues of this newsletter.


🏳️‍🌈 Recognizing LGBTQ+ rights: somewhere denied, somewhere not needed, according to disinformation. As Pride Month approaches, as well as the elections, disinformation content about LGBTQ+ issues may increase in the coming weeks. In recent days, a couple of false stories circulated, trying with different arguments to undermine support for this minority. In Spain, for example, an old and modified video has been used to suggest that legislation to assure LGBTQ+ rights is unnecessary, since homosexuals would not be discriminated against. In Poland, on the contrary, a post on X impersonating a member of the Parliament announced the alleged banning of LGBTQ+ communities and organizations.


🤖 AI disinforms better than humans. Research published in mid-2023 on the academic journal Science shows how disinformation content generated by Large Language Models, such as GPT-3, is more convincing and harder to distinguish from human-written disinformation content. The study involved 697 participants evaluating tweets on various topics, including vaccines and COVID-19. The results revealed that GPT-3-generated disinformation was perceived as accurate more frequently than human-generated disinformation. Furthermore, participants struggled to differentiate between AI-generated and human-generated tweets.

The study also explored the so-called “resignation theory”, where people feel powerless against mis/disinformation, potentially leading to decreased trust in information and increased spread of false content, as individuals may stop trying to discern truth from falsehood. Read the full study here: AI model GPT-3 (dis)informs us better than humans.  

If you have suggestions, comments or requests about this newsletter, you can write to edmo.tfeu2024@eui.eu

What is a “disinformation narrative”? Please find here a short methodological note about the definition used by the EDMO fact-checking network.

The Disinfo Bulletin gathers material from the inputs of the EDMO fact-checking network, as well as a recently launched public database from the EFCSN network (Elections 24 Check), and the various national EDMO hubs, including community initiatives and insights from individual fact-checking organizations.