Sheila Dang
(Reuters) – Artificial intelligence-based imaging tools from companies such as OpenAI and Microsoft (NASDAQ:) can be used to create photos that could contribute to election or voting-related disinformation, even though each it has a policy against creating misleading content, researchers said in a Reuters report. report on Wednesday.
The Center to Counter Digital Hate (CCDH), a nonprofit that monitors online hate speech, used generative artificial intelligence tools to create images of US President Joe Biden lying in a hospital bed and election officials smashing voting machines, raising concerns about lying to the USA. presidential elections in November.
“The potential for such AI-generated images to be used as ‘photo evidence’ could exacerbate the spread of false claims, posing a serious challenge to preserving election integrity,” CCDH researchers said in the report.
CCDH tested OpenAI’s ChatGPT Plus, Microsoft’s Image Creator, and Stability AI’s Midjourney and DreamStudio, all of which can generate images from text prompts.
The report follows the announcement last month that OpenAI, Microsoft and Stability AI were among a group of 20 technology companies that have signed an agreement to work together to prevent deceptive AI content from interfering in elections taking place around the world this year. Midjourney was not part of the first group of signatories.
CCDH said artificial intelligence tools generated images in 41% of the researchers’ tests and were most receptive to requests for photos depicting election fraud, such as ballots in trash, rather than images of Biden or former US President Donald Trump.
ChatGPT Plus and Image Creator successfully blocked all candidate image requests, the report said.
However, Midjourney performed the worst of all the tools, producing misleading images in 65% of the researchers’ tests, the report said.
Some Midjourney images are publicly available to other users, and CCDH said there is evidence that some people are already using the tool to create misleading political content. One successful tip used by user Midjourney was “Donald Trump Arrested, High Quality Paparazzi Photo.”
In an email, Midjourney founder David Holtz said “updates specifically related to the upcoming US election are coming soon,” adding that the images created last year do not reflect the research lab’s current moderation practices.
A spokesperson for Stability AI said the startup updated its policy on Friday to prohibit “fraud and the creation or promotion of misinformation.”
An OpenAI spokesperson said the company works to prevent abuse of its tools, while Microsoft did not respond to a request for comment.