The internet is drowning in AI-generated content. Deepfakes, voice clones, and fabricated images are no longer the stuff of dystopian sci-fi; they’re Tuesday afternoon. And governments, predictably, are responding the only way they know how: with fines.
From Spain to South Dakota, anti-deepfake legislation is gaining traction. Spain’s new bill threatens companies with penalties reaching $38.2 million or a percentage of global turnover for failing to label AI-generated content. South Dakota, not wanting to be left out, is considering civil and criminal penalties for those who spread deepfakes intended to sway political campaigns. It seems the future of misinformation might just be very, very expensive.
The EU AI Act, that sprawling regulatory behemoth, is the inspiration for much of this. Spain’s legislation seeks to enforce its transparency requirements, classifying deepfakes as “high risk.” Oscar Lopez, Spain’s Digital Transformation Minister, stated, “AI is a very powerful tool that can be used to improve our lives … or to spread misinformation.” He forgot the third, and increasingly popular option: generating mildly amusing cat videos.
South Dakota’s bill, on the other hand, is a bit more targeted. It focuses on political deepfakes shared within 90 days of an election, conveniently exempting newspapers, broadcasters, and radio stations—those bastions of unbiased truth. There’s also a carveout for satire and parody, which, let’s be honest, is wide enough to drive a fleet of AI-generated trucks through. Defining satire is notoriously difficult, leaving plenty of room for legal wrangling.
This patchwork of state laws in the US gained momentum after an AI-generated clone of President Biden’s voice urged New Hampshire voters to skip the primary. The FCC responded by slapping the alleged mastermind, Steve Kramer, with a $6 million fine. A deterrent? Perhaps. Or perhaps just a nice chunk of change for the government. Regardless, New Hampshire Attorney General John Formella declared that it sends “a strong deterrent signal.” Let’s hope Kramer had good lawyers, or at least a good sense of humor.
Beyond political trickery, the most prevalent form of harmful deepfakes is nonconsensual sexual content. Four states—Florida, Louisiana, Washington, and Mississippi—have already criminalized its distribution. One researcher estimated that over 244,000 deepfake porn videos were uploaded to major sites in a seven-year period, with a surge in recent months. Apparently, nothing says ‘progress’ like easily generated, digitally simulated indecency.
Even First Lady Melania Trump has weighed in, supporting the “Take It Down Act,” a federal bill targeting nonconsensual intimate imagery. If passed, platforms would be forced to remove such content within 48 hours. A noble goal, but as the Electronic Frontier Foundation (EFF) points out, these laws can be overly broad. They could be used to censor legitimate speech or incentivize false accusations. “Good intentions alone are not enough to make good policy,” warns EFF Senior Policy analyst Joe Mullin. Truer words were never typed.
So, what does the future hold? More fines, more legislation, and likely, more legal challenges. Tech companies and political campaigns with deep pockets will undoubtedly fight these laws, stretching government resources and potentially undermining their effectiveness. The question remains: can we curb the deepfake tide without drowning free speech in the process? Or will we simply create a lucrative new industry for lawyers and government coffers?
Leave a Reply