General Mills fights AI deepfakes on TikTok, Facebook and YouTube

General Mills wants to cut misinformation out of a balanced media diet. The food maker announced today a new partnership to combat it on social media platforms, including TikTok, YouTube and Facebook. General Mills will start measuring disinformation on social media sites by using Zefr, an independent measurement and verification firm, which will report how much misleading content the brand is up against when it runs campaigns online. The misinformation safeguards are part of a wider effort across the industry to apply new standards that prevent brands from appearing near unwanted categories of content. Related: Dove uses deepfake to warn of toxic influencers Misinformation represents a looming crisis online, especially considering the rise of generative AI, which has put deepfake technology into the hands of the masses. “It’s an unbelievably difficult problem to solve,” said Rich Raddon, Zefr’s co-founder and co-CEO. “The difference is how dangerous and how impactful disinformation and misinformation is going to be now, post generative AI.” Check out the latest Ad Age events and award programs This year, tools such as ChatGPT, DALL-E and Midjourney opened the floodgates on AI-generated voice-overs, videos and images. There have been seemingly playful uses of AI, such as a meme featuring Pope Francis in a puffy Balenciaga parka. There have also been instances where AI demonstrated how it could be used to concoct fake news, such as the AI images of ex-President Donald Trump getting handcuffed. The fakery is popping up all over social media, too. Gaming videos of President Joe Biden, Barack Obama and Trump are another popular genre. Making pictures of Trump getting arrested while waiting for Trump's arrest. pic.twitter.com/4D2QQfUpLZ— Eliot Higgins (@EliotHiggins) March 20, 2023 “If you think about all the things that come your way that aren’t actually accurate in this world of generative AI,” Raddon said, “it’s all very complicated for brands.” Read more: AI used to create fake brand collabs General Mills, the company behind Annie’s, Cheerios, Lucky Charms and Pillsbury, among other brands, is the first to apply misinformation reporting to its ad campaigns through Zefr, according to Lisa Roebuck, media director at General Mills. “We’re continually looking for ways to improve where our brands show up,” Roebuck said, “and to ensure that we're showing up in environments that are not supportive of explicit content or negative or content.” Last year, the Global Alliance for Responsible Media (GARM), an ad industry trade group, designated misinformation as one of the 12 categories of harmful content that brands should avoid as part of safety measures negotiated with major digital platforms. Other categories include violence, sex, obscenity, hate speech and spam, and GARM has been grading platforms on how well their tools perform remove content that breaks their policies. For content that falls within a platform’s policies, but outside of a brand’s sensibilities, marketers want to avoid running adjacent to that material. Meta, which owns Facebook and Instagram, recently launched its new adjacency control system it created with Zefr to sanitize feed environments. AI marketing blog How brands are using ChatGPT, DALL-E 2 and other tools Read more here Brands can set filters on their social media ad campaigns to steer clear of violating content such as misinformation. Zefr, so far, only issues post-campaign reports on TikTok to show General Mills an independent measurement on how well the misinformation filters performed, Raddon said. Facebook and YouTube are expected to offer the post-campaign reports soon, Raddon said. GARM has arranged its standards so that brands can choose “high,” “medium” and “low” risk settings for their ad buys depending on their comfort with certain topics. On misinformation, “high risk” involves “gratuitous” depictions of misinformation. “Medium risk” covers dramatic depictions of misinformation in an entertainment setting. “Low risk” covers news and articles that may discuss the subject of misinformation. Misinformation grey areas abound, though, with contention over the truth of news stories and coverage of current events. In fast-moving situations such as the coronavirus pandemic, some information is technically correct one day, but outdated the next. Last year, Zefr bought Israeli startup Adverif.ai to identify misinformation based on consensus drawn from fact-checking organizations. That work is going into the new misinformation reporting for brands. General Mills is using its partnership with Zefr to hold social media companies accountable if they don’t keep misinformation in check. “No misinformation is the standard,” Roebuck said. “We know that, entering the digital space, there always is going to be a level of risk.” General Mills works with Mindshare, the global media agency within GroupM. Brands are worried about misinformation, in general, but there also are direct threats to the brands, where people on social media are posting misinformation about their products. There have been high-profile cases of top brands becoming the center of social media storms based on faulty information. One such case: In 2020, Wayfair was the target of far-out, baseless conspiracy theories involving human trafficking, which circulated on social media. More recently, oat milk brand Oatly has pushed back against what it deems misinformation about its ingredients, and the company made fact-checking a centerpiece of its marketing. McDonald’s has combatted misinformation about its ingredients for years, refuting persistent talk of “pink slime,” which stems from the way McDonald’s used to make hamburgers but doesn’t anymore, according to an AP fact-checking article last year. Coca-Cola has dispelled myths about cocaine ever being an ingredient in its soft drink. Brands are worried about “the reputational risk of brand association with misinformation and the social and political risk of bad actors controlling the narrative,” said Joe Barone, managing partner, brand safety Americas, GroupM.

General Mills fights AI deepfakes on TikTok, Facebook and YouTube

Featured Stories