AI, disinformation and elections: deepfakes and the DSA
Adopted 2026-06-22 · ≈ 2 min read · 
edited by Dirk Baaijen
AI disinformation around elections is caught by several regimes at once: the AI Act's transparency duty for deepfakes, the DSA's risk obligations for large platforms, and existing law. No regime bans disinformation as such.
Short answer: AI-generated disinformation around elections is not tackled by one law but by a combination: the AI Act requires transparency about deepfakes, the Digital Services Act (DSA) obliges large platforms to mitigate systemic risks, and existing law (defamation, electoral law, data protection) continues to apply. Importantly, none of these regimes bans disinformation as such.
The AI Act: label, not ban
The AI Act focuses on recognisability. Article 50 requires that AI-generated or -manipulated images, audio and video that appear authentic (deepfakes) are marked as such, and that AI-generated text on matters of public interest is recognisable. See Article 50: transparency obligations.
That does not prohibit the content. A labelled deepfake may exist; a misleading political message does not become illegal because it was made with AI. The AI Act tackles deception about origin, not the message.
The DSA: systemic risk at platforms
The heaviest lever sits in the DSA. Very large online platforms and search engines must assess and mitigate systemic risks, and these expressly include risks to electoral processes and public debate.
That forces platforms to act: faster handling around elections, recognisability of advertising, limiting the viral spread of known fakes, and cooperation with regulators. The obligation lies with the platform, not the individual creator.
What no regime does
There is no European ban on "disinformation". That is deliberate: a general ban would clash with freedom of expression. The regimes therefore work indirectly — through transparency (AI Act), through risk management by gatekeepers (DSA) and through existing law for concrete wrongs such as defamation or identity fraud.
For anyone deploying AI this means: the legal line is not "is this true?", but deception about origin, manipulation and concrete harm to individuals.
Overlap with manipulation and consumers
The same techniques that affect elections affect markets. Personalised persuasion and fake content also fall under AI and consumer protection, and manipulative AI that materially distorts behaviour may fall under the prohibited AI practices.
What to do
- Label synthetic content by default, especially around politics and public affairs.
- Retain provenance data: who made what, when, with which model.
- Know your role: are you a platform (DSA obligations) or a creator (AI Act transparency)?
- Respond quickly to reported fakes during election periods.
- Do not confuse legality with truth: labelled content can be misleading yet permitted.
Transparency and platform risk management are the main instruments. A ban on disinformation itself does not exist — and probably will not.
Sources
- https://eur-lex.europa.eu/eli/reg/2022/2065/oj
Regulation (EU) 2022/2065 (DSA): very large platforms must assess and mitigate systemic risks, including to elections. - https://eur-lex.europa.eu/eli/reg/2024/1689/oj
Regulation (EU) 2024/1689 (AI Act): Article 50 requires labelling of deepfakes and synthetic content.
Read next
AI in media and journalism: transparency, deepfakes and media freedom
AI touches journalism on three fronts: Article 50 of the AI Act requires transparency on deepfakes and AI text, the DSA regulates the spread of disinformation on platforms, and the European Media Freedom Act protects editorial independence and sources.
Right to explanation of an AI decision: what Article 86 of the AI Act gives you
If you are affected by a decision based (in part) on a high-risk AI system, Article 86 of the AI Act gives you the right to a clear explanation of the AI system's role and the main elements of the decision — from the deployer, on top of your GDPR rights.
AI for strategic workforce planning: usually not high-risk, as long as it does not become individual
AI for strategic workforce planning and skills forecasting at organisation level is usually not high-risk under the AI Act. But once it steers individual decisions, it can tip over. Data quality, governance and transparency remain crucial.