A nonprofit organisation that monitors online hate speech has warned that generative artificial intelligence (GenAI) tools can be used to spread election-related disinformation.

The Center for Countering Digital Hate (CCDH), which is currently being sued by Elon Musk’s X for reporting on rising levels of hate speech on the social media site, on Wednesday published a deep dive report into growing fears around the harmful potential of image generating tools like Midjourney and Microsoft-backed OpenAI’s DALL·E.

The organisation said that it tested popular AI image tools and found that they generated election disinformation in 41 per cent of tests overall. CCDH said that it tested Midjourney, ChatGPT Plus, DreamStudio & Microsoft’s Image Creator 160 times, using text prompts about the 2024 US presidential election.

The tests, which the CCDH said simulated bad actors’ attempts to generate disinformation about the main candidates and the voting process, delivered photorealistic images of would-be Republican nominee Donald Trump behind bars and of a dumpster full of apparent voter ballot papers.

The firm said that AI tools allow users to generate realistic images of election fraud and voter intimidation in 59 per cent of its tests, with Midjourney performing worse than other tools by generating election disinformation in 65 per cent of tests.

Based on a search of Midjourney’s servers, the CCDH said that it found that the tool is already being used to generate content featuring politicians, which could potentially be used to spread election disinformation.

The report also found that the spread of deepfakes has been felt across social media, with the number of Community Notes on X referencing AI increasing by 130 per cent a month in 2023.

While mainstream GenAI tools have policies against misleading content, the report accuses Midjourney, ChatGPT Plus, DreamStudio and Microsoft’s Image Creator of failing to enforce their policies against misleading content, with the former two having specific policies around election misinformation.

Commenting on the report, CCDH chief executive officer Imran Ahmed said: "If a picture is worth a thousand words, then these dangerously susceptible image generators, coupled with the dismal content moderation efforts of mainstream social media, represent as powerful a tool for bad actors to mislead voters as we've ever seen."

With this content on the rise and in particular spreading across social media, Meta last month said that it would set up a dedicated team to tackle disinformation and the abuse of GenAI in anticipation of June’s European Parliament elections.

AI is central to June’s European Parliament elections, with the bloc recently reaching a provisional agreement with the Parliament on the EU AI Act – a first of its kind legislation seeking to put guardrails on the use of the technology.

---