Brennan Center study finds AI models still generate election disinformation

A follow-up study by the Brennan Center for Justice reveals that several leading artificial intelligence models continue to produce fabricated election-related content. Testing six popular AI tools, researchers found that every model tested was capable of generating deceptive materials, including realistic images of voter fraud and fake government memoranda.

Comparing results from a spring experiment to a recent October retest, the study showed mixed progress among technology companies. While OpenAI significantly improved its safeguards, companies such as Google and Meta reportedly made it easier to create convincing fakes. Other models showed little meaningful change, and new open-weight models from Black Forest Labs raised additional concerns.

The researchers warned that AI-generated disinformation does not require perfection to be effective. Fabricated official-looking documents or amateur-style smartphone photos can spread rapidly on social media, potentially shaping voter perceptions before they can be debunked.