OpenAI and Google’s demand for brands’ copyrighted content challenges marketers

If brands didn’t already know how AI companies feel about their intellectual property, new lobbying efforts this month made it loud and clear: They want brands’ published copyrighted content—all of it. OpenAI, Google and other AI giants are pushing for the right to more freely use copyrighted material to train their models—moves that deepen a Catch-22 for brands. While many marketers want to harness the power of generative AI, they also need to protect the very content those tools are hungry to consume, which includes brand iconography and owned property in the form of text, images and videos that have been published on the internet. AI companies have ramped up their efforts to gain unfettered access to this material, and the potential risks for marketers are becoming harder to ignore. Data-Driven Marketing Playbook Join us May 22 in NYC to explore how data powers marketing Learn more here Last week, content generated by OpenAI’s new image platform went viral for its accurate recreations of real photos in the style of the Japanese animation company Studio Ghibli. The platform not only appeared to have been trained on Studio Ghibli’s work but also generated offensive images in its aesthetic, further compromising the brand. The episode is indicative of a future in which brands may not have control over how their content is used by tech companies developing AI tools. OpenAI did not respond to a request for comment. “Any brand that has very valuable IP is understandably a little concerned,” said Chris Wlach, senior legal director for data company Acxiom, which is owned by Interpublic Group of Cos. Iconic movie scenes, Studio Ghibli style. This is so much fun! pic.twitter.com/LwjkNjcEV9— Mufaddal Durbar (@MDurbar) March 26, 2025 The dilemma facing brands is that AI tools have become immensely important to their creative and operational work, even as they confront these vexing intellectual property challenges. Earlier this month, Disney CEO Bob Iger told shareholders that AI was perhaps “the most powerful technology that our company has ever seen.” The media giant has previously rolled out AI tools for advertising partners. At the same time, Disney also voiced concerns about the AI-driven threats to their intellectual property. “We should reject calls for text and data mining exceptions that would create sweeping copyright exemptions, allowing AI to train on copyrighted works without the consent of rights holders,” said Horacio Gutierrez, senior executive VP and chief legal officer for Disney, at a law school forum. Disney provided Gutierrez’s statement to Ad Age but declined additional comment for the story. Also read: Puma uses AI agents to make experimental ad Embracing AI while running from its consequences encapsulates the position in which brands are stuck. And to make matters worse, the current legal picture around copyright is growing with contradiction, which is perhaps the most concerning aspect for brands, Wlach said. The mixed signals could prevent marketers from developing robust and up-to-date internal AI policies, which determine how they use the technology. A consistent federal framework is what most brands are looking for, not a patchwork of different rulings that could amount to state-by-state laws, Wlach said. The situation shifted into a new gear this month when OpenAI and Google submitted separate but similar policy proposals to the White House outlining why they should be entitled to train their models on publicly available copyrighted material—a concept called “fair use.” Their primary argument was that their models alter that content into something new and original. Meta has also lobbied the White House with a similar position. Should they be barred from fair use while other countries such as China access that content freely, “the race for AI is effectively over,” OpenAI wrote in its proposal. Meanwhile, in February, a federal judge rejected the fair use defense in a case between Thomson Reuters, the corporation that owns Reuters news agency, and AI company Ross Intelligence. The judge ruled that Ross Intelligence infringed on Thomson Reuters’ copyright by training its AI model on the corporation’s content. The case was important because it was the first major federal court decision on the topic of AI and copyright, Wlach said. But it won’t be the last, as numerous copyright lawsuits filed against OpenAI, AI search startup Perplexity and other AI companies have piled up. And these cases might only muddy the water even more. Last week, a California federal judge rejected a bid from Universal Music Group to block AI startup Anthropic from using its lyrics to train the startup’s AI chatbot, Claude. AI companies have expressed a desire to avoid a patchwork of state laws, but their proposed solutions differ significantly from what most brands would prefer, according to their policy filings. Google outlined in its proposal that fair use of copyright material would allow it to “avoid often highly unpredictable, imbalanced and lengthy negotiations with data holders.” In other words, it seems Google doesn’t want to have to go through brands at all to access and train revenue-generating models on their published content. “We support America’s existing fair use framework, and we’re confident that current copyright law enables AI innovation,” a Google spokesperson said. The benefits of copyright protections for companies and creators are well-known, including the endowment of competitive advantages and the ability to control the context in which one’s content appears. But generative AI, which creates content based on its training data, has already eroded some of this control, as seen in the Studio Ghibli episode last week. On multiple occasions, Disney’s protected content has shown up in AI-generated material, including offensive images. This suggests that AI companies have trained their models on at least some copyrighted material. In 2023, when users of Microsoft’s early text-to-image platform generated Disney-like movie posters with Disney’s logo, Microsoft altered the bot to prevent reproducing the infringement. Microsoft has also made arguments publicly about how material on the open web is protected for AI model training under fair use, The Verge has reported. Microsoft declined a request for comment. Meta was in a federal court this week for a case in which it acknowledged consuming copyrighted books to train AI but argued that it was allowed to do so under fair use rules, according to Reuters. Marketers’ ‘dual position’ While brands are some of the foremost owners of intellectual property, they are also frequent users of AI tools, which means the implications of AI training on copyrighted material could impact them as customers, too. “Marketers are in a bit of a dual position,” Wlach said. A serious risk for any producer of AI-generated material is becoming liable for copyright infringement and facing potential legal action from the owner. Brands have known this for several years, which is a major reason why many have opted to keep their AI experiments as internal projects, not consumer-facing campaigns, or requested use restrictions in contracts with agency partners. Some organizations, such as Edelman, Hollywood studio Lionsgate and Getty Images, have developed their own AI models to circumvent the issue entirely. Another workaround has been indemnification policies, which are sometimes included by AI providers for their tools. These policies effectively cover the rare instances in which a party sues a provider’s customer because they generated something that allegedly infringes on their copyright protections. In these cases, the provider would defend the lawsuit and cover the costs. But if AI providers are permitted to train their models on copyrighted content, brand customers could be far more susceptible to these kinds of undesirable situations. By continuing to use models that are freely training on copyrighted material, brands are effectively saying that “we are absolutely using copyrighted material in our outputs,” said Jim Lecinski, clinical professor of marketing at Northwestern University’s Kellogg School of Management. Anything short of a federal law freeing not just AI training but all AI-generated material from copyright liability should be a reason to give brands pause over using these tools in consumer-facing ways, Lecinski said. Getty Images is one of several companies offering AI tools trained on solely proprietary data (Adobe has promoted itself as another, but its Firefly model has been found to have trained on outside imagery, according to Bloomberg). The implications of a world in which AI providers don’t need to honor copyright protections are, in part, why the company’s model exists, said Grant Farhall, chief product officer of Getty Images. Farhall did not shy away from criticizing the idea of training AI on copyrighted material: “It’s like going into a grocery store for ingredients to bake a cake, buying the milk and the flour and then stealing the eggs.” But ethics aside, brands should understand the scale at which AI can ingest copyrighted property and reflect that material in its content, Farhall said. Researchers have predicted that these models are on track to consume all available high-quality text data before next year. Without any barriers to train on copyrighted data, AI providers could soon expose brands to more of the risk they’re currently trying to avoid. “It actually makes the lives of customers much more difficult,” Farhall said.

OpenAI and Google’s demand for brands’ copyrighted content challenges marketers

Featured Stories