COLUMBUS, Ohio — Ohio lawmakers are trying to combat misinformation by requiring disclaimers on realistic artificial intelligence-generated or created content.
“AI — every single day, [it] seems to be getting better and better," state Rep. Adam Mathews (R-Lebanon) said.
And it’s becoming increasingly accessible. There are now widely accessible, free, or inexpensive services that use AI to create and edit photos, videos, and audio. But with that, Case Western Reserve University technology law professor Erman Ayday says come major problems.
"People can also use [sites] in order to create effects for themselves, like you can actually put yourself in front of a beach right now," Ayday said. "But sometimes, they're also used for manipulation, for bad purposes, for blackmailing."
Over the past several years, deepfakes have become more prevalent online — with fake explicit photos of celebrities to politicians urging you not to vote for them.
RELATED: Proposed Ohio bill would require disclaimers for political AI deepfakes, add criminal penalties
Mathews is trying to stop this type of misinformation. He and state Rep. Ty Mathews (R-Findlay) proposed House Bill 185, which would regulate AI content.
"This is a way for the normal Ohioans to protect themselves," A. Mathews said.
Malicious deepfakes, AI meant to harm someone’s image, would only be allowed as long as they have a clear watermark that they were artificially manipulated.
"If it would confuse a reasonable person, there is some type of disclaimer," the lawmaker added. "If a reasonable person would say that this is a manipulated image, then that would not need a disclaimer."
Let’s show some examples.
Original photo
This is the original photo of Rep. Mathews by a sign for now Vice President JD Vance.
Deepfake 1 — Needs disclaimer
This is a deepfake I made of him — one by a Kamala Harris sign. This image would require a disclaimer.
Deepfake 2 — Doesn't need disclaimer
This is an expertly drawn "deepfake" I made of him. This one would not require a disclaimer.
"Political cartoons are totally protected; no one thinks that a drawing actually happened," Mathews said.
Enforcement
The legislation would allow everyone to own their image, and it would be a trademark infringement to produce malicious content without someone’s consent. There could be civil penalties of tens of thousands of dollars.
The bill also makes it a third-degree felony to create or distribute malicious AI content in order to extort the individual being deepfaked.
Any pornographic deepfakes or deepfakes of a child are completely banned, with or without a disclaimer.
No one has spoken out publicly against the bill, but Ayday said the biggest challenge will be enforcement.
"This is not really something scary for people who are doing this for malicious purposes — especially during the elections or during some high-stakes events," the professor said. "Distributing this thing in an anonymous way is extremely easy."
The bill will continue being heard this week.
