This is part of a recurring series of stories aimed at educating readers on how to break into advertising and marketing, or land a new position. Read the last installment here.

Generative AI is the shiny new toy many advertisers and marketers are eager to play with, but using it poses several legal risks that are still being ironed out. Given the uncertainty, agencies and brands are weighing the risk of using these tools prematurely—and potentially getting themselves sued—against the risk of not experimenting at all and being left in the dust.

Copyright and IP ownership are two of the most pressing issues, with pending court cases that have been brought against generative AI developers over the data scraped from the internet to train their various systems. The New York Times, for example, is suing ChatGPT maker OpenAI for using its articles to train its apps.

The outcome of such lawsuits will likely determine whether generative AI apps are, by design, infringing on third-party copyrights, and who exactly has rights to their IP in this new space.

Advertising trade bodies, consultants, lawyers and newly appointed chief AI officers are all trying to figure out exactly how they can experiment with generative AI in marketing in the least risky ways possible while keeping up with the various new developments and litigation that are constantly surfacing.

Also: AI in ads divides consumers—what brands should know

Ad Age spoke with multiple to understand the safest ways to experiment with generative AI as a marketer or advertiser. Outlined below are some of the key issues and tips to help.

Appoint someone to spearhead AI use

A good first step to using generative AI in advertising is appointing one person at your agency or within your marketing team to look into the ways it can be safely used, according to several of the experts including lawyers who spoke with Ad Age for this story.

There are many resources advertisers and marketers can lean on to keep up with the developments in generative AI including alerts from the Federal Trade Commission (for those in the U.S.) and law firms, said Katie Gardner, partner at law firm Gunderson Dettmer. Since the technology is evolving so quickly, Gardner suggested agencies and marketers appoint one person whose full-time job is to understand how the rest of the company can implement and experiment with generative AI.

“They really need a person or group for keeping up with what’s happening and making sure that they’re staying on top of any changes,” Gardner said.

Several agencies have already hired people in such a role. Accenture Song, for example, recently hired Arun Kumar to lead the agency’s team focused on devising solutions around data and AI, as its parent company invests $3 billion into AI over the next couple of years.

Nick Watts, partner at creative production agency Hook, took on the role of chief AI officer at the beginning of 2024 to understand the possible use cases for generative AI at his company. He told Ad Age that part of his job has been presenting a monthly report for the agency. “I’ll do tutorials. I’ll show stuff off in some departmental meetings … I distill it down to one hour a month,” he said.

Watts outlined his routine for keeping up with the developments in the space: “A morning cup of coffee and scouring the internet in different places.” He said social media can be a good resource. “Even [reading through] Reddit and X [formerly Twitter], finding the right people to stay connected to because it all just gets posted out there,” he said.

Also read: Top 5 AI recent marketing activations

Use generative AI for internal use

The biggest concern about potential copyright infringement with generative AI in advertising focuses on the output, several of the experts and lawyers said. That means an agency, marketer or generative AI developer is most likely to get sued if something that is copyrighted ends up in an AI-created ad that is public. And who exactly is liable for that is also still up in the air at the moment.

“The huge issue is how these tools and models have been trained in the first place,” said Roch Glowacki, managing associate on the ​​digital, commerce and creative team at law firm Lewis Silkin. Did the data “come from legitimate sources or not? If you are an advertiser and your marketing team is working on something or you’re engaging an agency and they are using these generative AI tools, there’s a question of whether the content that it spits out is potentially infringing third-party rights.”

This is why multiple ad industry insiders said it’s best to use generative AI tools for internal matters, including strategy planning within a marketing team or for sparking ideas for creative campaigns at an agency.

Related: 9 AI tools agencies are embracing right now

“There’s a stage in this sort of development process where the agency or the marketer should be able to play around with these tools to just, at least, create new ideas they might not necessarily [think of],” Glowacki said. “What sometimes ends up happening is, you give [the generative AI app] 10 different prompts, you see what it has generated and then you get an 11th idea, which is actually your idea, and you go off and then you shoot it yourself or you get the right people in the room to deliver it.”

When litigation started arising, many marketers told their employees and vendors, including agencies, they couldn’t use generative AI tools at all, according to Simon Francis, founder and executive chairman of consultancy Flock Associates. That’s still the case for some companies but his firm has since put together a risk assessment to help brands and agencies use generative AI tools in safe ways as they don’t seem to be going anywhere anytime soon and the greater risk may be in not using them at all.

“People started to engage with us, [saying] ‘Hey I’m caught between a rock and a hard place,” Francis said. “My lawyers are saying I can’t do anything risky. I’ve got marketers wanting to experiment. I’ve got agencies wanting to make money and also terrified because they think [generative AI] is going to eat their breakfast; they’re trying to do something with it to show that it can add value rather than reduce value. How can we go into this space to make sense of it?”

One option Francis advises is for marketers and agencies to train generative AI models on only their owned assets, so, think past campaigns and images that came from the brand itself, with closed-source AI tools. Some apps including ChatGPT offer AI services specifically for businesses, so companies pay subscriptions so they can use the tools in a more secure, private space.

“From the legal perspective, the internal use and gen AI running on closed or proprietary data poses relatively minimal risk; you have rights to the data, it’s your data,” said Dave Ervin, co-chair of law firm Crowell & Moring’s advertising and brand protection group.

Still, there are certain risks to assess including how comfortable the marketer feels sharing that data with its agency partners who are likely going to use it to make the ad, Francis said. He advised those fine details to be worked out in contracts.

Get it in writing

Experts advised every possible restriction or use of generative AI in advertising to be laid out in writing: in marketer and agency contracts; contracts with influencers; contracts with production companies, etc.

Brands have started introducing AI restrictions to agency contracts, which Ad Age previously reported has caused some tension between shops that are racing to adopt generative AI, and clients who are worried about all of the ways the technology could get them in legal trouble.

“In the negotiation of agency agreements, clients are asking and requiring disclosures for the use of gen AI in provisions of scope of services,” said Ervin, who negotiates agreements between brands and agencies. “‘Did these six concepts come from a gen AI model? I just want to know.’”

Ervin said agencies need to disclose how they’re using generative AI in their work “at a minimum.”

Disclosures will come into play in the actual ads created, too. If an ad was partially created using AI, that likely needs to be disclosed to the public viewing the ad, experts including lawyers interviewed said.

“What will consumers be exposed to?” said Gabrielle Robitaille, associate director of digital policy for the World Federation of Advertisers (WFA), which is one trade organization that has put together an AI community to understand the impact of generative AI on advertising. “And with that, [comes] the debate around transparency for consumers. Do we have to be disclosing when we use generative AI content in our assets? What’s going to be best practice? Are we watermarking? Are we labeling? How is that going to look?”

Glowacki said he could see fashion or beauty brands using generative AI to create product ads but they have to consider if the generative AI tools “polished” the product too much in a way that it would become deceptive marketing. “You might be able to very easily exaggerate the qualities of a tanning product or something else,” and that may violate certain advertising rules, he said.

Ervin said he’s also helping advertisers and marketers go back into old contracts for campaigns that involved various parties including ad directors to determine whether there is anything in them that would suggest they couldn’t train their own proprietary generative AI models on those campaigns.

“We didn’t know enough to get permission for gen AI models because they didn’t exist five years ago,” he said. “If we knew about it, we would have asked the producer. Does it fall in the scope of rights that fell into the agreement?”

The process has been “very time-consuming,” Ervin said, and hasn’t yet yielded a clear answer.

What tools should you use?

The WFA AI community is made up of 600 senior marketers, Robitaille said. “Part of that group is trying to understand how can we leverage generative AI to drive marketing efficiencies and effectiveness while mitigating legal, reputational and ethical risks?”

That WFA group provides information “necessary about risks associated with generative [AI] so brands can make their own, informed decisions about which tools to use,” Robitaille said. The tools that marketers allow their agencies to use when it comes to their campaigns should be explicitly written in contracts.

Several experts said Adobe’s AI-powered image generation and expansion tool, Firefly, is a good one because Adobe has said it will help companies, including by paying damages, if they are sued for copyright infringement over any assets its tool creates.

Adobe is “standing behind the output,” Gardner said. “If you're using something like that, then you can rely on some protection.”

Hook’s Watts said his agency uses Adobe Firefly to “help get our creative professionals closer to the creative output you [as the client] want.”

How best to use generative AI for external matters

While most legal experts advise advertisers and marketers to only use generative AI tools for internal use, brands such as Toys R Us and Under Armour have come out with ads partially created with generative AI. There was some industry backlash and criticism over their quality, but that’s another debate.

Generative AI is here and it’s possible that many future ads could be created with the help of these tools.

For now, one way an advertiser could use generative AI to create an ad in the safest way possible is by using still images that do not show people or use music that could otherwise be copyrighted, Ervin said. It might be used for product shots or as part of a larger campaign. “[That use] could have the potential to be cost-efficient … that’s a big attraction,” he said.

Another big risk marketers need to be aware of if they are using generative AI to help create an ad, Glowacki said, is they may give up ownership of that ad altogether, or relinquish that ad, subject to the relevant terms of services, to the generative AI developers themselves.

That discussion is being had in China currently, he said. There, the debate is around if the prompt the human enters into the generative AI tool is “sophisticated” enough, the company or person entering that prompt may be able to retain ownership. This underscores the need for human intervention.

“So, if I type into these generative AI systems that I want an ad for a lawyer and it’s as simple as that, and it spits out a picture of a lawyer, that's not original enough to be protected by copyright,” Glowacki said. “But if the prompting was much more sophisticated … then that prompt might be sophisticated enough to meet that originality threshold.”

“It would be an awkward situation for a company to have an ad that it cannot actually protect by copyright,” he added.

Due to these types of risks, multiple ad experts advise marketers and agencies to avoid using generative AI in the creation of ads, at least until legal matters are sorted out.

“Gen AI output is so risky,” Ervin said. “Advertising is a really competitive world. I work with some really cool brands that are putting a lot of sweat, tears and resources into this stuff. At the end of the day, [you can’t have] your ownership and control of those materials diminished or non-existent."