AI-generated political memes rack up Facebook likes

Driving the news: The Center for Countering Digital Hate released a report Thursday detailing their findings from studying 169 different AI-generated posts on Facebook.

• Each of the posts featured an image of a fake person, including those posing as military veterans, sharing right-leaning political views.
• These posts were shared more than 476,000 times between July 1 and Oct. 23.

Zoom in: In one post that was still on Facebook on Wednesday, a military veteran is seen holding a folded American flag with a message that reads, "Veterans deserve better than being second to student loans."

• That image had roughly 54,000 likes and 6,800 shares as of Wednesday.
• In another post that has been removed, an elderly female veteran is shown with the message, "They'll hate me for this, but learning English should be a requirement for U.S. citizenship!" The post had more than 128,000 reactions and the initial wave of comments came from users who said they agreed with the post.
• However, the image of the elderly veteran failed to accurately replicate actual military ribbons, per the report. They weren't lined up properly, and the colors didn't correspond to any real military honors.

The intrigue: Meta, which owns Facebook and Instagram never promised to to label all AI content, but they have said they'll label what their systems detected as AI-generated.

• "We will begin adding 'AI info' labels to a wider range of video, audio and image content when we detect industry standard AI image indicators or when people disclose that they're uploading AI-generated content," the company said in July.
• Meta's policies toward labeling have evolved as AI-generated images have appeared on its network.
• During their study, CCDH's researchers could not identify a clear way to report AI-generated content.
• Researchers tested that these posts were AI-generated using tools from Hive and asking other researchers to verify their findings.

The big picture: Meta's annual developers conference last month was filled with announcements about the company's own AI generated content plans, including tests of "content imagined for you by Meta AI that will appear in your Facebook and Instagram Feeds."

Yes, but: Many of the posts linked in the center's report have already been removed.

• Users have flooded the comments on other posts to warn that the content is AI generated, noting that pins on uniforms aren't right or that a flag is folded incorrectly.
• It's worth noting that according to Meta's earnings report this week the company currently has 3.29 billion users across Facebook, Instagram and WhatsApp, so 2 million likes is a small percentage of that number.

What they're saying: "User-generated corrections, they come far too late to deal with the vast majority of people who've been exposed to the disinformation," Imran Ahmed, CEO of the Center for Countering Digital Hate, told Axios.

• "This really comes down to the AI platforms' failure to come together effectively to protect our democracy by setting standards for transparency and accountability," he added.

Threat level: Some pages sharing this content had administrators based in Morocco, Pakistan and Indonesia, researchers found.

• "What we are seeing is there are certain jurisdictions — some of them we've identified here in Morocco and Pakistan — that are becoming the global arms dealer in a race for more and more devastating generative AI disinformation," Ahmed said.

The other side: Meta did not immediately respond to a request for comment.

Go deeper: AI is already making it easier to spread election lies