The report also revealed that these AI platforms could create election disinformation in 41% of their test runs. Midjourney was found to be most likely to produce potentially misleading results. While all platforms had some rules against using their tools to mislead others, the challenge lies in enforcing these rules. The CCDH has urged AI companies to collaborate with researchers to test and prevent 'jailbreaking' and encouraged social media platforms to invest in identifying and preventing the spread of potentially misleading AI-generated images.
Key takeaways:
- AI image generators can be manipulated to create misleading election-related images, raising concerns about the spread of political misinformation, according to a report by the Center for Countering Digital Hate.
- Despite pledges from leading AI firms to address these risks, the report suggests that some companies still have work to do to ensure their AI tools cannot be manipulated to create misleading images.
- AI platforms Midjourney, Stability AI’s DreamStudio, OpenAI’s ChatGPT Plus and Microsoft ImageCreator were tested and found to be capable of creating misleading images related to US presidential candidates or voting security.
- The report found that the AI image generators created election disinformation in 41% of their test runs, with Midjourney most likely to produce potentially misleading results.