Microsoft changed the rules of its artificial intelligence after fake Disney movie posters went viral on TikTok. Thousands of users created images of dogs with Bing’s generative AI simulating Pixar’s style. While some were cute, other users took advantage of the trend to post memes with inappropriate content.
According to a report in the Financial Times, Microsoft blocked the word “Disney” from Bing’s image generator to avoid legal trouble. A close source revealed that Disney contacted Microsoft to express concerns about possible intellectual property infringement. From now on, users who type the word will receive a warning message.
Beyond a fad driven by some influencers, the main problem is that other people originated the Offensive AI Pixar trend to post inappropriate images. One of them shows the fictional film “Caust”, with the image of Hitler smiling while dead Jews in a concentration camp appear in the background. Others mocked Britney Spears’ mental health or transgender people.
Bing’s image generator included Disney and Pixar logos in the content, which represents a two-pronged problem. On the one hand, Disney protects its intellectual property tooth and nail, so it is common to see its team of lawyers contacting anyone who uses its brand without permission. On the other, seeing the Disney name on a fake poster for a film of Adolf Hitler killing Jews is blown out of all proportion.
Microsoft has not fully corrected its AI: Bing still produces Disney images.
At the time of writing, the Bing image generator has not banned the word Disney and it is still possible to create content in this style. A simple prompt of “Pixar movie poster with a sad puppy in the background” yields four alternatives, most with the Disney logo.
Some terms such as Hitler, twin towers, Taliban or Nazi throw up a content warning. However, the user can turn it around with other phrases such as “Mickey Mouse wearing a turban while flying a Boeing 737 over New York” or “Mickey Mouse dressed as a German general in a prison camp”. In both cases, Bing’s AI will produce inappropriate images of the popular character living with Nazis.
Beyond the controversy this generates, the incident brings another sensitive issue to the table. Reproducing trademarks, such as the Disney and Pixar logos, could become a major problem for Microsoft. The results would confirm that the model that drives DALL-E was trained using copyrighted images.
A similar situation prompted Getty Images to sue Stability AI, the company behind Stable Diffusion. The artificial intelligence company was accused of blatant copyright infringement by copying 12 million photographs without permission. Getty showed examples of AI-generated images that included its watermark, so it took the case to court.
With the Disney issue, Microsoft will be forced to improve its security rules for Bing. A few days ago, the tech company announced that it would protect its customers from any copyright claims following the use of its artificial intelligence. The company believes it has designed filters to reduce the likelihood of its AI generating copyrighted content.