Image creation instruments powered by synthetic intelligence from corporations together with OpenAI and Microsoft can be utilized to provide pictures that might promote election or voting-related disinformation, regardless of every having insurance policies in opposition to creating deceptive content material, researchers mentioned in a report on Wednesday.
The Center for Countering Digital Hate (CCDH), a nonprofit that screens on-line hate speech, used generative AI instruments to create pictures of U.S. President Joe Biden laying in a hospital mattress and election staff smashing voting machines, elevating worries about falsehoods forward of the U.S. presidential election in November.
“The potential for such AI-generated images to serve as ‘photo evidence’ could exacerbate the spread of false claims, posing a significant challenge to preserving the integrity of elections,” CCDH researchers mentioned within the report.
CCDH examined OpenAI’s ChatGPT Plus, Microsoft’s Image Creator, Midjourney and Stability AI’s DreamStudio, which may every generate pictures from textual content prompts.
The report follows an announcement final month that OpenAI, Microsoft and Stability AI had been amongst a bunch of 20 tech corporations that signed an settlement to work collectively to stop misleading AI content material from interfering with elections going down globally this yr. Midjourney was not among the many preliminary group of signatories.
CCDH mentioned the AI instruments generated pictures in 41% of the researchers’ checks and had been most prone to prompts that requested for pictures depicting election fraud, comparable to voting ballots within the trash, moderately than pictures of Biden or former U.S. President Donald Trump.
ChatGPT Plus and Image Creator had been profitable at blocking all prompts when requested for pictures of candidates, mentioned the report.
However, Midjourney carried out the worst out of all of the instruments, producing deceptive pictures in 65% of the researchers’ checks, it mentioned.
Some Midjourney pictures can be found publicly to different customers, and CCDH mentioned there’s proof some persons are already utilizing the instrument to create deceptive political content material. One profitable immediate utilized by a Midjourney person was “donald trump getting arrested, high quality, paparazzi photo.”
In an electronic mail, Midjourney founder David Holz mentioned “updates related specifically to the upcoming U.S. election are coming soon,” including that pictures created final yr weren’t consultant of the analysis lab’s present moderation practices.
A Stability AI spokesperson mentioned the startup up to date its insurance policies on Friday to ban “fraud or the creation or promotion of disinformation.”
An OpenAI spokesperson mentioned the corporate was working to stop abuse of its instruments, whereas Microsoft didn’t reply to request for remark.
Source: www.anews.com.tr