AI-generated slopaganda in political ads puts internet platforms on alert

The rise of slopaganda

The term "slopaganda" describes a new phenomenon where generative AI "slop" is combined with propaganda and misinformation. This trend has become a significant challenge for major internet platforms like Instagram, YouTube, and X as they try to monitor content. Experts first used the term early last year to describe this specific type of deceptive media.

A recent analysis by the Wesleyan Media Project revealed that more than $187 million has been spent on nearly 300 ads that were either generated or enhanced by AI. The report suggests this number is likely an undercount because many of these ads do not include any disclaimers. This lack of transparency makes it harder for the public to know when they are viewing synthetic media.

The spending on these ads is not evenly distributed across the political spectrum. According to the report, over 90% of the spending came from Republican candidates, a Republican party committee, or a Republican-leaning group. This concentration of spending highlights how quickly AI tools are being integrated into modern campaign strategies.

Ads appearing without labels

Users are increasingly encountering AI-generated ads and posts alongside authentic videos from politicians. For example, CNBC found a fake video of New York Democratic Governor Kathy Hochul posted by her Republican challenger, Bruce Blakeman. This content appeared online without any clear disclosure to the viewer.

In another instance, Republican senatorial candidate Ken Paxton created an AI-generated video of his Democratic opponent, James Talarico, making controversial statements. The video included a three-second disclosure flashing in the middle, but it was shared as organic content on YouTube and X by the campaign. These examples show the varying ways candidates choose to present synthetic media to the public.

The lack of consistent labeling creates a fragmented experience for social media users. While some campaigns use brief flashes of text, others provide no indication at all that the content is artificial. This inconsistency makes it difficult for platforms to enforce a uniform standard for what constitutes a "disclosure."

Platform responses and rules

Different tech companies have adopted different strategies for managing AI content. YouTube told CNBC that it applies "prominent labels to realistic AI content" and allows political candidates to request the removal of videos where their faces are altered. For paid advertisements, the company requires disclosure when synthetic people or events are used to provide transparency.

Meta has also ramped up its efforts by adding an "about this ad" tool for content detected as being created with third-party AI tools. The company maintains an ad library and mandates "paid for by" labels for all political ads. However, Meta also places specific restrictions on the use of these ads during the week prior to an election.

X takes a different approach by stating that advertisers must comply with applicable laws. This puts the primary responsibility on the advertiser rather than the platform itself. These varying rules mean that the level of protection or information a user sees depends heavily on which app they are using.

Testing AI safety and bias

Researchers have been testing how popular AI models handle requests for deceptive content. The Brennan Center for Justice tested models like ChatGPT, Gemini, and Claude to see if they would facilitate misinformation. Their August study found that while many models avoid familiar conspiracies, they can often be manipulated into creating misleading material.

A follow-up test in October showed that some progress has been made by specific models. For instance, ChatGPT began redirecting prompts for harmful election images into illustrations that look like comics. It also started refusing to take action on certain prompts altogether to prevent the generation of deceptive content.

Despite these improvements, the risk remains that generative AI can be used to create highly convincing fake images and videos. The ability of these models to be manipulated means that platforms must constantly update their detection tools. This ongoing arms race between content creators and safety filters defines the current state of AI in politics.

Proposed federal legislation

Several national politicians are currently trying to address the issue of AI-generated misinformation through new laws. One proposal, the Protect Elections from Deceptive AI Act, is backed by senators from both parties. This bill would prohibit the knowing distribution of materially deceptive AI-generated video or audio about federal candidates.

Other bills are being led by Democrats in Congress, such as the AI Ads Act. This legislation aims to expand existing bans on campaign misrepresentation to specifically include generative AI content. These efforts represent a move toward creating a federal standard for AI transparency in the political arena.

As of now, none of these proposed bills have advanced out of committee. While there are currently about 30 state laws with different requirements for AI disclosure, candidates do not always abide by them. The lack of a comprehensive federal law leaves a significant gap in how AI-generated political content is regulated nationwide.