“Slopaganda”: AI political ads flood the US midterm elections

According to CNBC, the US midterm election campaign has been dominated by so-called "slopaganda": campaign ads featuring clips and images generated by artificial intelligence. This poses a new challenge for platforms such as Instagram, YouTube and X as they try to police their spaces. The term was coined early last year, when a group of AI experts used it in an academic paper to describe a phenomenon that combines generative AI "slop" (mass-produced, low-quality content) with propaganda and disinformation.

The legal framework is weak. There is no comprehensive federal law regulating the use of AI in political ads or requiring its disclosure. There are some 30 state laws, with differing requirements on the use and disclosure of AI, but candidates do not always comply with them. On Meta, Google and X platforms, users encounter AI-generated ads and posts from politicians mixed in with authentic videos and clips.

The Wesleyan Media Project analysis, published last week, counts more than $187 million spent on nearly 300 ads generated or enhanced with AI. The organization warns that the figure is almost certainly an undercount, because many ads carry no disclaimers and it is sometimes hard to determine whether AI was used. More than 90% of the spending came from Republican candidates, a Republican party committee or an allied group. Most of the ads did not include disclaimers about AI generation.

CNBC found several AI ads without disclaimers. One is a fake video of New York's Democratic governor, Kathy Hochul, posted by her Republican rival, Bruce Blakeman. In Texas, Republican Senate candidate Ken Paxton created an AI video in which his Democratic opponent, James Talarico, makes controversial statements. It includes a disclaimer that flashes for three seconds midway through the video, and the Paxton campaign is distributing it as organic content on YouTube and X. The Blakeman and Paxton campaigns did not immediately respond to requests for comment.

The platforms' responses vary. YouTube says it applies "prominent labels" to realistic AI-generated content. For organic content, candidates can request the removal of material in which their face has been altered or generated by AI. It requires advertisers to disclose when an election ad contains synthetic or digitally altered people or events. Google also requires political ads to pass its election ads verification and to include a "paid for by" disclaimer. X maintains that advertisers must comply with applicable laws, which places responsibility on them rather than on the platform, and did not respond to CNBC. Meta pointed to its political ad rules, which include restrictions in the week before elections, "paid for by" labels and a political ad library. In June it strengthened labeling with an "about this ad" tool for ads it detects as created with third-party AI tools.

Another risk for tech companies is being asked to create misleading images and videos. The Brennan Center for Justice tested ChatGPT, Gemini, Grok, Claude, Perplexity and DeepSeek to see whether they would facilitate misleading content. Its August study concluded that many models avoid generating content about well-known conspiracies, but can generally be manipulated into producing misleading material. A new test in October showed progress in ChatGPT, which began redirecting requests for harmful election images toward comic-style illustrations, or refusing them.

There are initiatives in Congress. The Protect Elections from Deceptive AI Act, backed by senators from both parties, would ban knowingly distributing materially deceptive AI-generated video or audio about federal candidates. Democrats are pushing other bills, such as the AI Ads Act, which would extend existing bans on campaign misrepresentation to generative AI content. None of these bills has made it out of committee.

Note: the downloaded text also includes background information on Meta, including a $25 billion fine that a New Mexico court reportedly imposed on it last month in a privacy case linked to Cambridge Analytica. That figure comes solely from the article and I have not verified it through other means.