A video that Lt. Gov. Jay Collins reposted on X in late June depicts a dystopian scene in which the Florida Capitol burns, robots arrest citizens, and a man asks gubernatorial candidate Byron Donalds to shut down data centers because of a lack of clean water. In the clip, Donalds replies, “Get this peasant out of my office,” while Collins, dressed as the superhero Captain America, violently attacks symbols of data centers, abortion clinics and Donalds himself. The clip was created using generative artificial intelligence and contains no disclaimer required by Florida law.

Florida’s statutes mandate that any AI‑generated content that is knowingly false and intended to deceive voters must prominently display a notice stating it was created with generative AI. The video shared by Collins does not include such a disclaimer, raising questions about compliance with the state’s transparency rules.

The incident follows a study by the Media and Journalism Research Center that examined 40,000 posts on LinkedIn and Facebook. According to the study, 90 % of those posts were either partially or entirely AI‑generated. The research highlights how AI is increasingly embedded in social‑media feeds, from the algorithms that curate content to the synthetic media that populates user timelines.

Alex Mahadevan, director of MediaWise at the Poynter Institute, said that a few years ago it was easier to spot deepfakes by looking for visual clues such as missing fingers or out‑of‑focus teeth. “Now it is becoming essentially impossible to tell the difference between an AI‑generated deepfake of a politician or an actual video of the politician speaking,” Mahadevan told reporters. He added that the “information ecosystem” is in a state of flux and that viewers must treat online content with skepticism.

Joshua Scacco, who leads the Center for Sustainable Democracy at the University of South Florida, noted that negative attack ads are not new. “In the age of artificial intelligence, candidates and campaigns don’t necessarily need to rely only on that moment; they can create essentially a false moment and market it as if it had happened,” Scacco said. He pointed out that Florida’s law, while intended to increase transparency, does not help users identify which parts of an ad were generated by AI.

A Pew Research Center survey released earlier this year found that half of adult Americans express concern about the rapid rise of AI. The same study suggests that many voters are unsure how to distinguish authentic content from synthetic media.

Marbin Pazos Revilla, a cybersecurity expert at USF, warned that the speed of AI development creates loopholes in existing regulations. “AI is advancing so rapidly… it’s hard not to have loopholes,” Revilla said. He urged lawmakers to tighten restrictions on AI‑generated political content.

The video’s creator has not been identified, and Collins declined to comment when asked for a statement. The clip was produced by WUSF as part of a statewide journalism initiative funded by the Corporation for Public Broadcasting.

Fact‑checking tips

Experts recommend several steps to verify the authenticity of online videos:

Use reverse image search tools such as Google Images, TinEye, or InVid. Zoom in on details—look for inconsistencies in horizons or button shapes. Search reputable news outlets for coverage of the same event. If a clip appears too perfect or aligns too closely with a viewer’s beliefs, treat it with caution.

These methods can help voters discern whether a piece of media is genuine or a product of generative AI.

The incident underscores the growing challenge of AI‑generated political content and the need for clearer regulatory guidance. As Florida’s gubernatorial race heats up, the use of synthetic media may become a pivotal issue for both candidates and voters.

The story was produced by WUSF as part of a statewide journalism initiative funded by the Corporation for Public Broadcasting.