AI-generated photographs of animals undermine the credibility of real content on social media

Wired describes the case of a woman named Mibbby Butler, whose missing cat Brooklyn was reportedly found in April. A stranger sent her a photograph of the cat and demanded money for its temporary care before returning it. Butler noticed that the cat in the photo was in the same pose as on her own missing-cat flyer and that the text on a label in the background was blurred and illegible, a typical sign of an AI-generated image. She did not pay the money; according to the article, the cat remains missing four months after the incident, and Butler receives similar scam messages roughly once a month.

According to the article, the nonprofit organization We Animals faces a similar problem. It publishes photographs and videos by 175 photojournalists documenting the treatment of animals on farms, in circuses and in laboratories. Although the organization prohibits its photographers from using AI, some social media users still accuse it of using AI because the images (for example, from a calf-rearing operation on a farm in Arizona) look almost too perfect. According to its representatives, the organization plans to start sharing behind-the-scenes footage and introduce technology that records the origin and editing history of files to maintain audience trust.

According to philosopher and activist Oscar Horta, creators of AI slop profit from likes and advertising revenue, while cheap synthetic videos, he says, are overwhelming real content in social feeds and search results. Horta fears that fabricated videos of animal rescues during fires or floods will lead people to question legitimate rescue practices and make it harder to raise funds for these activities. He also highlights the risk that similar false content (for example, deepfakes of drowning polar bears being rescued by people on boats) could inspire dangerous behavior toward animals.

As a response, the article mentions that Jeff Sebo from the Center for Mind, Ethics, and Policy at New York University called on AI developers this year to include an emphasis on ensuring that generated content does not harm animals in their model rules. According to the article, new laws in California and the EU now require the most widely used image generators to embed invisible markers revealing the AI origin of the content, which social networks are supposed to use to label such content. Content verification tools are available in ChatGPT, Gemini and Meta AI, but they have limitations, such as a limit on the number of images that can be checked.