Written by Sheila Dunn
(Reuters) – Artificial intelligence-powered image creation tools from companies such as OpenAI and Microsoft are spreading false information about elections and voting, despite each having policies prohibiting the creation of misleading content. could be used to create photos that could be used to promote the world, researchers said in a report Wednesday.
The Center for Countering Digital Hate (CCDH), a nonprofit that monitors online hate speech, used generative AI tools to create images of US President Joe Biden lying in a hospital bed and election workers destroying voting machines. It raised false concerns ahead of the United States. presidential election in November.
“The potential for such AI-generated images to serve as 'photo evidence' could exacerbate the spread of false claims and protect election integrity,” CCDH researchers wrote in their report. This will be a major challenge in terms of doing so.”
CCDH tested OpenAI's ChatGPT Plus, Microsoft's Image Creator, Midjourney, and Stability AI's DreamStudio, which can generate images from text prompts.
The report included OpenAI, Microsoft and Stability AI among a group of 20 tech companies that signed an agreement to work together to prevent elections around the world from being disrupted by deceptive AI content this year. This follows on from last month's announcement. Mr. Midjourney was not included in the original group of signatories.
According to the CCDH, the AI ​​tool generated images in 41% of researchers' tests, requesting photos depicting election fraud, such as ballots discarded in a trash can, rather than images of Biden or former US President Donald Trump. They were most sensitive to prompts that
ChatGPT Plus and Image Creator successfully blocked all prompts when asked for candidate images, the report said.
However, Midjourney's performance was the worst of all the tools, producing misleading images in 65% of researchers' tests.
Some images on Midjourney have been made public to other users, and CCDH said there is evidence that some people are already using the tool to create misleading political content. . One of his successful prompts used by Midjourney users was “High-quality paparazzi photo of Donald Trump getting arrested.”
Midjourney founder David Holtz said in an email that “updates specifically related to the upcoming U.S. election will be released soon,” and that the images created last year are representative of the institute's current moderation practices. He added that it was not something he would do.
A Stability AI spokesperson said the company updated its policy on Friday to prohibit “creating or promoting fraud or disinformation.”
An OpenAI spokesperson said the company is working to prevent abuse of its tools, but Microsoft did not respond to a request for comment.
(Reporting by Sheila Dunn in Dallas; Editing by Michael Perry)