How might advertisers be impacted by California’s new AI watermarking bill?

“Failing to appropriately label synthetic content created by GenAI technologies can skew election results, enable defamation and erode trust in the online information ecosystem,” the bill’s text reads. By enacting the bill, “the Legislature can help to ensure that Californians remain safe and informed.”

Introduced earlier this year by Assemblymember Buffy Wicks, AB 3211 arrives at a time of mounting social and political tension around deepfakes, or AI-generated content that’s realistic enough to trick the casual observer into believing that it's authentic. Some brands have been quick to embrace deepfakes in their marketing efforts. They're also being wielded in the political arena in an effort to sway voters, though not so much as many had feared (at least not yet).

Other brands, meanwhile, have become concerned about the role of AI in advertising. In April, marking the 20th anniversary of its ‘Real Beauty’ campaign, beauty brand Dove pledged to never to use AI-generated depictions of women in its ads.

California’s new bill is indicative of a cultural trend in the direction of transparency around AI, and marketing professionals would do well to jump aboard, says Alon Yamin, CEO of AI governance platform Copyleaks. “For California-based companies and especially any marketing teams utilizing AI, this means adapting strategies to prioritize authenticity and build trust with audiences,” he says. “It’s time we all start to embrace transparency and ensure that AI is used responsibly and ethically in marketing campaigns and everywhere else.”

As it stands now, Deepfakes are not an especially common sight in advertising, but that could change as generative AI tools continue to improve and multiply. If passed, AB 3211 would mean that AI-generated ads could be identifiable to California-based viewers, either through embedded watermarks or visible labels. This could have adverse effects for the brand not unlike the way that a self-identifying AI-generated robocall tends to prompt most people to hang up in just a few seconds.

“For brands and marketing agencies using generative AI, they should be aware that when they display advertisements containing AI-generated content, prospective customers may start to see labels indicating that the advertisement does not reflect a real-person, object, place ... or event,” says Arsen Kourinian, an attorney specializing in AI governance and data privacy at law firm Mayer Brown. “Advertising agencies should consider what type of chilling effect this may have on clicks and conversion rates.”

The rise of deepfakes, and the ease with which they can be created, has raised a quagmire of legal and ethical dilemmas. The AI industry is almost completely unregulated in the US, which means the matter of labeling has fallen mostly to the companies building AI systems. The situation is a bit different in the European Union: its recently enacted AI Act includes a clause mandating the disclosure of AI-generated content.

California is, of course, the cradle of the American tech industry, which makes it an especially significant locale for this legal action around government-enforced watermarking.

Adobe, which has previously implemented its own labeling policies for AI-generated content, has thrown its support behind the bill. The company “worked together with Assemblymember Wicks to provide technical expertise on how to leverage provenance technology to bring this vision to life,” Anne Perkins, Adobe's director of policy and government relations, said in a statement. “Today’s bill provides an effective framework that allows good actors to be trusted, while protecting people from harmful AI-generated deepfakes, which is especially critical in an election year when misinformation tends to swirl and the stakes are high. We applaud Assemblymember Wicks for driving this critical piece of legislation that allows AI innovation to thrive while protecting the public against the misuse of this technology.”

Tech giants Microsoft and OpenAI are also onboard with the new bill, according to a Monday report from TechCrunch. (Those companies could not immediately be reached for comment.)

By legally requiring the disclosure of synthetic content, AB 3211 could make Californians more amenable to the use of AI, says Deborah Perry Piscione, author and CEO of Work3, a Bay Area-based consulting firm which helps businesses embrace emerging technologies. The bill “has the potential to build greater public trust in AI technologies and digital platforms,” she says. “By increasing transparency and authenticity in online interactions, it could help maintain the integrity of personal and brand communications in an increasingly AI-driven world. This trust could, in turn, encourage further innovation as users become more comfortable with AI-generated content that is clearly labeled as such.”

Assemblymember Wicks was also the author of the California Journalism Preservation Act, another bill introduced in the state earlier this year that sought to force tech companies to pay a portion of their ad revenue to news outlets based in the state. That bill culminated in an agreement between state officials and tech companies, which will establish a fund for California news publishers, along with a program, designed to equip organizations across a range of industries with new AI tools. Google has pledged to pay $30m in the first year and at least $20m annually over the next two to five years as part of the deal.

California lawmakers are also currently deliberating SB 1047, a bill that would require companies building powerful AI models to adhere to certain safety protocols, at the risk of hefty fines.

AB 3211 passed the California senate appropriations committee earlier this month. If it’s approved by the full state Senate by the end of this month, it will advance to the desk of Governor Gavin Newsom, who can either veto it or sign it into law.

For more on the latest happenings in AI and other cutting-edge technologies, sign up for The Emerging Tech Briefing newsletter.