You see a candidate reach for his phone. He looks perfectly real. The voice sounds exactly like him. But the man on the screen and the words from his mouth are completely fabricated. This is a deepfake political ad. Its target was California’s Democratic gubernatorial candidate Xavier Becerra. The creator was Republican strategist Steve Hilton.
This is not science fiction. It is the reality of the 2024 election cycle. From local school board races to congressional contests, AI-generated slopaganda is flooding the political landscape. This new breed of content blends propaganda with artificial intelligence slop. Its aim is not always to deceive. Often, it seeks to ridicule, to satirize, to shock. It operates in a gray area where free speech protections meet digital manipulation.
Jonathan Underland knows this firsthand. He is a campaign staffer for Becerra. He also became an unwilling character in Hilton’s viral deepfake ad. “It’s moving so fast that I don’t think anything really, truly shocks me anymore,” Underland said. He described a process of grim acceptance. You absorb the hit and roll with the punches. Then he saw another ad. This one depicted a deepfake Becerra in a migrant detention center. The AI candidate ordered a woman to be terminated. Police threw her into a closet.
Underland had not seen that video before our interview. His reaction was visceral. “It’s really infuriating to see that,” he said. “I feel like I have to sit with that for a little bit. That is beyond the pale.” He noted the manipulative details. The AI had darkened the candidate’s skin. It made his hair appear greasy and sweaty. These are not accidental choices. They are classic tools of the attack ad, now supercharged by AI.
California is now trying to draw a line. The state recently updated its laws on digitally altered political content. The new rules require clear labeling of AI-generated videos, images, or audio. The restrictions become more stringent in the 120 days before an election. “The next best step is to think about how we actually enforce this,” said Stefani Langehennig. She tracks deepfake political ads at the University of Denver’s Center for Analytics and Innovation with Data. “Can we take down these ads, or are we stepping on somebody’s freedom of speech?”
Enforcement remains an open question. Langehennig is unaware of any deepfake political ad being removed solely for legal violations. Most states have now passed laws regulating this content. Yet the ads continue to spread. Hilton’s videos, for instance, include a disclosure. They state the content is AI-generated. When asked if the ads mislead voters, Hilton defended their narrative. “No, they’re telling the truth,” he asserted. “People need to know that this guy, Becerra, has got no plans to solve any of California’s problems.”
- Backlash from voters
- Manipulated video impact
- Public trust issues
- AI’s role in misinformation
- Legal enforcement challenges
- AI in political campaigns
The impact on voters is still unclear. Experts say we lack the data to understand how these manipulated videos sway public opinion. We do have early examples of backlash. In a San Francisco race to replace Nancy Pelosi, candidate Scott Weiner deployed a parody AI chatbot of his opponent Connie Chan. The stunt provoked significant criticism. Weiner later acknowledged it “missed the mark.”
| Issue | Description |
|---|---|
| Deepfake Ads | AI-generated political content that can mislead voters |
| California Law | New regulations on labeling of AI-generated content |
| Public Trust | Concerns about the erosion of trust in political candidates |
| Manipulation | Use of AI to enhance negative perceptions |
| Satire vs Deception | Gray area between comedic content and harmful misinformation |
| Future Implications | Potential consequences for elections and voter perception |
For Jonathan Underland, the path forward is troubling. He would never use such tactics for his own clients. He believes this technology actively destroys the public’s trust. “Now there’s so many interfering signals,” he explained. “You can’t tell what’s real and what’s not about a candidate. That seems like we are headed down a dark path.”
The core challenge remains. How do you regulate a tool that can be used for satire, for parody, and for malicious deception? California’s labeling law is a first step. It is an attempt to provide viewers with a crucial signal. This content is manufactured. But a label does not stop the emotional reaction. It does not undo the vivid, false memory an ad can plant. The law sets a boundary, but the battle for truth in politics has entered a new, deeply synthetic phase. The real test will be whether voters can learn to distrust their own eyes.
