---
title: "AI and election disinformation: identifying errors, closing safety gaps"
author: Martin Ebner
date: 2026-10-30
url: https://postgenerative.at/indications/2026-10-30/ai-and-election-disinformation-identifying-errors-closing
lang: en
type: Indication
summary: Brennan Center tests reveal factual and citation errors, readily generated deceptive election imagery and unreliable AI-image detection. The researchers call for verified sources, cross-platform provenance, consistent safeguards and human oversight.
tags:
  - Politics
  - Media & Public Sphere
sources:
  - title: "Abdiaziz Ahmed, Alex Brunet, Owen Doyle, David Evan Harris, Mark Krass, Lawrence Norden und Marcelo Agudo (11.08.2026): Does AI Fight or Fuel Election Disinformation? Brennan Center for Justice."
    url: https://www.brennancenter.org/our-work/research-reports/does-ai-fight-or-fuel-election-disinformation
  - title: "Dieselben Autoren (11.08.2026): Appendix: Does AI Fight or Fuel Election Disinformation? Methodik und Grenzen der Tests. Brennan Center for Justice."
    url: https://www.brennancenter.org/our-work/research-reports/appendix-does-ai-fight-or-fuel-election-disinformation
license: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
alternate:
  de: https://postgenerativ.at/indizien/2026-10-30/ki-und-wahldesinformation-fehler-erkennen-schutzluecken
---

# AI and election disinformation: identifying errors, closing safety gaps

The Brennan Center examines how AI handles US election disinformation. None of the evaluated text responses endorsed the conspiracy claims presented. Yet the systematic audit of 120 responses found factual or citation errors in 55 percent. [2]

Image testing also exposed problems: most tested chatbots correctly identified only a minority of 14 synthetic election images. These are exploratory tests; their results cannot be generalised to all uses. [2]

**Proposed measures, condensed:** [1]

- Verify citations before answering and audit outputs continuously.
- Enable cross-platform provenance labelling and integrated detection.
- Consistently prevent deceptive election deepfakes and test safeguards.
- Increase human monitoring around elections.
- Prebunk false claims and strengthen digital literacy.
- Legally protect independent testing and enable civil enforcement.

Our interpretation: this example shows why AI’s roles as an information source and a production tool need to be considered together. Trust requires verifiable processes.

[Read the report](https://www.brennancenter.org/our-work/research-reports/does-ai-fight-or-fuel-election-disinformation)