By Sam Bradley • October 17, 2025 •
Ivy Liu
OpenAI’s launch of Sora 2, its video generation tool and would-be social network for the slop age, was the latest attempt by the generative AI developer to expand the copyright Overton Window.
Sora 2 initially shipped with an “opt-out” policy for IP rightsholders. If an artist or advertiser didn’t want their work to be used in the tool’s model, they would need to get in touch. It wasn’t a popular move. “Early returns from rightsholders have suggested that they are not opting in,” said Glenn Pudelka, a copyright and privacy attorney at Troutman Pepper Locke, told Digiday.
Within days, OpenAI founder Sam Altman rolled back the policy in a blog post following criticism from publishers, movie studios and artists — shifting to an opt-in approach.
With over 30 AI-related copyright lawsuits currently in train, generative AI boosters have zeroed-in on copyright law as a roadblock in the way of the otherwise inevitable AI revolution. It’s why the U.K. government opened a formal consultation on whether it should relax IP rules in favor of tech companies hungry for large language model input data earlier this year. It’s a flashpoint for criticism of the tech, too. In June, U.S. Senator Josh Hawley (R-Mo.) referred to LLMs’ dredging of the web as “the largest intellectual property theft in American history.”
For advertisers, the episode highlighted agencies’ growing role as guides for brands navigating the most significant challenges to copyright — to who it is designed by, and who it’s designed to protect — since the advent of the music recording industry.
“It’s uncharted territory,” said Vicky Brown, general counsel commercial and chief privacy officer at WPP.
Prior to 2022, a client might only hear from an agency’s in-house counsel on matters of inspiration and homage — brought in, for example, to judge whether a TV spot stepped over another rightsholders’ intellectual property lines. (Ask your local creative director and odds are, they’ll tell you a war story about a great ad gutted due to legal jitters).
But since AI’s Cambrian explosion kicked off, they’ve become crucial players on agencies’ client services bench.
“Our legal team has definitely become more of a client services partner because of AI,” said Wesley ter Haar, chief AI officer at Monks.
While not every company has an in-house counsel like WPP’s Brown – both a trained lawyer and an agency veteran of 14 years – agency execs are having more conversations around legal advice. With AI proficiency a table-stakes matter in media and creative pitches, legal guidance relating to AI applications has also become a necessary and important part of client service work.
Day-to-day, that might mean advising clients worried about getting caught up in the crossfire of lawsuits brought by publishers, artists and authors against companies like OpenAI. Indemnification promises have become standard among agencies. Or it could include internal risk assessments for agencies sizing up a potential AI tech partner or startup — making sure they don’t get into bed with the wrong outfit in their quest to prove they’re handier with the tech than the shop around the block.
In the last fortnight, they’ll have been outlining each agency’s take on the Sora 2 copyright situation, even if they don’t operate in the U.S. (which at the time of writing is the only country Sora’s beta is available).
“Our clients are very exercised about ensuring that copyright is protected,” said one agency exec based in the U.S., who exchanged anonymity for candor.
Though major advertisers are often risk-averse, this isn’t just the usual slow walking. Marketers’ first worry is that their own IP – a mascot or logo in which they’ve imbued brand equity – might be misused.
“As you move closer to a world of artificial general intelligence, we’re going to be confronted more frequently by things that are human-like and in many ways, interpreted by people as being human,” said the anonymous agency exec.
If users are competing to depict Nintendo’s Kirby committing a terrorist act, or OpenAI founder Sam Altman cooking and eating Pikachu, there’s a risk that images showing Green Giant or the Hamburglar running amok might also flood the web. (You’d hope folks could be trusted to distinguish between satire and advertising, but Facebook users also thought those images of a crustacean Christ were real.) What threat might that pose to a brand’s carefully curated, family-friendly reputation?
Brand-safety hypotheticals aside, there’s also the concern that LLMs are crossing copyright lines regarding advertisers’ own IP. “Sora 2 has highlighted the fact that there can be a risk of copyright infringement in the models,” said Brown.
In each case, the ultimate question remains the same. By opting in, a brand could encourage the use of its IP in a grand feast of user-generated content. Let a thousand flowers bloom, even as their meadows are bulldozed for data centers.
And by opting out, a brand reserves the right to control its own destiny — while paying an opportunity cost. “Most brands would kill for high quality, user generated content engagement,” said ter Haar, while conceding the risks are high. “People are a little freaked out about the fidelity and the ease of [production] because that brings up more potential risk [of] things going viral, that you really don’t want to see go viral.”