Multiple states have passed laws requiring disclosures on artificial intelligence-generated deepfakes. Now one New York lawmaker wants to label content that is real.
Assemblymember Alex Bores (D) plans to introduce a bundle of bills next year to spur content authenticity labeling of images, videos and other media. It is an idea he hopes catches fire in other states.
“It is easier to prove what is real than to detect what is fake,” said Bores, who says he is the first Democrat with a computer science degree to be elected to a seat in Albany.
States have already begun grappling with the implications of generative AI and the ease with which people can now make synthetic media. More than half of states have passed laws to address election-related or pornographic deepfakes, according to tracking by Public Citizen.
California lawmakers passed Sen. Josh Becker’s (D) first-in-the-nation law requiring large generative AI companies to include a detection tool with their products, so that the public can see if content has been created or manipulated by AI. Florida lawmakers ordered an evaluation of digital provenance standards for images and audio created by generative AI.
Bores, who led the passage of an election deepfake law this year, intends to take a different approach for 2025. He has embraced an emerging global provenance standard known as C2PA that software maker Adobe and multiple other companies including Amazon, Google, Meta, Microsoft and OpenAI have coalesced behind. The Department of Defense is also an early adopter.
Like HTTPS, which is a secure web transmission standard, C2PA is an open-source technical specification that gets embedded in media content so that anyone can know its origins and how it has been edited or manipulated.
“What intrigues me is it’s the first time that I have felt like there is a solution to deepfakes that is not this cat and mouse game,” Bores said, referring to the challenge of identifying AI-generated content.
Bores told Pluribus News that he plans to introduce four C2PA bills for New York’s 2025 legislative session.
- A requirement that social media companies preserve C2PA metadata when someone uploads an image or video to their platform.
- A mandate that government agencies use C2PA authentication on any media they create.
- A directive that political campaigns, beginning with statewide campaigns in 2027, apply C2PA to their media content.
- A requirement that generative AI image generators tag their images with C2PA to identify AI-generated content.
A fifth bill Bores is weighing but does not plan to introduce next year would require that all hardware such as cameras and smartphones sold in New York have C2PA built in.
Bores said states have a role to play in jumpstarting C2PA’s widespread adoption, which is why he is also discussing the concept with lawmakers in other states.
“That’s pretty important because for this to be successful you have to get to a point where 90%-95% of images have this standard,” Bores said.
C2PA is rapidly gaining traction with tech companies, social media platforms, major news media outlets and camera makers. Adobe’s five-year-old Content Authenticity Initiative, which is promoting C2PA worldwide, claims more than 3,700 members.
Unlike a traditional watermark that is visible, C2PA resides in the background and relies on a combination of cryptographic metadata, watermarking and content fingerprinting to create a “durable content credential.”
Adobe describes C2PA as a “nutrition label” for digital media content that can be applied to audio, video, photographs and even documents to create a “digital chain of trust and authenticity.” Typically, a user would access the provenance data by clicking on a content credential icon in the corner of the image or media.
“We’re encouraged to see recent legislative proposals focusing on provenance, which will help bring this vision to life and ultimately build the trust and transparency we need in an AI-powered world,” an Adobe spokesperson said in a statement to Pluribus News.
A new report from the Future of Privacy Forum, a Washington, D.C.-based think tank, states that watermarking and other methods that address the rise of synthetic content can help users make sense of what they are encountering online. But the report also warns that data provenance requirements can expose sensitive personal information that puts individual privacy at risk.
“Lawmakers and organizations seeking to implement effective safeguards while also protecting privacy should consider these tensions when developing strategies for addressing harmful synthetic content,” wrote the report’s author, senior policy analyst Jameson Spivack.
Bores said he too has concerns about privacy and plans to carve out personal information from what is required to be preserved by his C2PA legislation.
Industry often fights tech regulation at the state level. But Bores is optimistic he can win support for adopting C2PA standards in New York because it is quickly emerging as an industry standard.
“I don’t think anyone opposes this,” Bores said. “It’s just getting the details right.”