Generative AI sparks brand safety concerns marketers know all too well

“Anything’s fair game,” said Ryan Meegan, co-founder and CMO at Dude Wipes, a flushable wipes brand. “You can’t really control that.” Anyone with an AI image generation tool subscription could replicate a brand campaign.

But to not use AI is to get left behind. In turn, marketers are choosing to deal with any repercussions regarding using AI if and when they come. It’s not that marketers don’t care. It’s that they’ve seen this show before, according to the six marketers Digiday spoke with for this story.

Even without AI, social media has always been a brand safety cesspool

“In many respects, this is similar to the early days of social media, when people would talk about brands in those spaces — both positively and negatively,” said Brian Yamada, chief innovation officer at VML.

Even without generative AI, brands have grappled with maintaining control of the brand narrative amid the rise of social media, user generated content, influencers and content creators. In some cases, a brand’s reputation was more impacted by influencer sentiments than official communications from the brand, said David Corns, chief commercial officer at Opendoor, a real estate tech company. AI is just the new frontier, he added. 

“Just as with UGC, the most successful brands will be those that use generative AI as an integral part of their ecosystem and leverage it to strengthen their connection with consumers,” Corns said. 

What do marketers actually control?

Brands’ battle to control their narrative in the cultural zeitgeist is nothing new. Even before generative AI, social media — or the internet for that matter — brands have grappled with their brand logos and marketing messaging taking a life of their own in the public stratosphere, beyond their direct control.

Recall Mastercard’s “Priceless” campaign from the 90s. The campaign was replicated and imitated by everyone from Saturday Night Live, mocking the sex scandal involving former U.S. President Bill Clinton and Monica Lewinsky, a White House intern, to Ralph Nader, presidential candidate in 2000, in a campaign spot targeting his opponent George Bush. (Mastercard sued Nader for appropriating the corporate ad.)

More recently, film company A24 spoofed the “Priceless” campaign to promote its new movie “Materialists” last month. To Mastercard, it’s a matter of earned media — as long as it paints the brand in a positive light, according to Cheryl Guerin, evp of brand strategy and innovation at Mastercard. “I would say it extends our brand and our visibility. It might be OK if it’s being done in a right way,” she said.

Marketers test out brand safety remedies for AI

The pressure to do more marketing with fewer marketing dollars has nudged the industry toward AI. One agency exec, who spoke on the condition of anonymity, said the agency is pushing internal teams to create custom agents for client brands, using brand style guides, past campaign decks and other IP to build out agents and train AI models.

The agency operates in what the exec calls a “zero-trust architecture,” otherwise known as “the principle of “never trust, always verify.” While staff is encouraged to use the agency’s proprietary AI tools, there are guardrails and monitoring systems in place to ensure client information is protected, via pop ups warning staff of responsible use, legal team training and monitoring of company machines.

“We are able to audit what everybody’s doing in the sense of what outputs are coming out, what inputs are going in, so that we can track [AI use],” per the exec. Still, there’s only so much an agency or brand can control, including how staff use AI-powered tools on their personal accounts and devices.

Notably, AI chatbots like ChatGPT use user input to train models and improve the chatbot’s performance — making it crucial to understand what inputs are going into those models to ensure it’s not shared with other chatbot users outside of the company or agency. Generative models, especially those built on an enterprise and commercial level protect brand IP, according to Yamada. But, as noted, those aren’t the only tools on the market and safeguards aren’t guarantees.

Brands could issue cease-and-desists, but legal actions could be limited with little government or industry oversight. AI copyright infringement hasn’t made it to the courts yet, but it has become a talking point for some brands and companies, said David Teske, partner in Alston & Bird law firm’s technology and privacy group. It’s an issue that has put the social media companies distributing AI-generated content on the frontlines: AI videos have taken over social media feeds, leaving platforms like Meta, YouTube and TikTok grappling with how to protect user trust.

Some brands, like Coca-Cola, have started experimenting with AI-generated ads. Per Teske, content created solely by AI isn’t protected by copyright law because law protection requires a human creator. In other words, brands using AI to generate content could be copied and reused by others without permission.

“You do have risks with brands that output is generated from generative AI is not protected under law and can be replicated and reused by others,” Teske told Digiday.

As AI adoption continues, its promise of productivity, speed and cost savings still outweigh concerns about impacts on climate change, measurement to calculate marketing spend efficiency and systematic errors. Just tack IP and brand safety onto the growing list of bridges marketers will cross when they get there. 

“AI is rapidly shaping our future. It’s the new frontier. Consider it a co-pilot [to advertising], and be ready and willing to adapt,” said Corns.