Meta: Labeling AI-generated images

Meta has announced its intention to label AI-generated images and videos on its social media platforms Facebook, Instagram and Threads.

Meta President Nick Clegg: "We have labeled photorealistic images created with Meta AI right from the start so that users can recognize how they were created."
Corresponding images are labeled "Imagined with AI". In order to also label AI images created with third-party tools, Meta has worked with partners on common technical standards that signal when content was created with AI.
Invisible markers in accordance with the C2PA and IPTC standards will be used to label images from Google, OpenAI, Microsoft, Adobe, Midjourney and Shutterstock so that other platforms can also recognize them.
Meta also wants to oblige users to label their AI-generated content accordingly on their own initiative. Those who do not comply will be sanctioned. "But it's not yet possible to identify all AI-generated content, and there are ways to remove invisible tags," says Nick Clegg. "We are working hard to make it more difficult to remove or alter invisible watermarks." Meta's AI research lab FAIR, for example, has developed a technology called Stable Signature Watermarks that cannot be deactivated."People and organizations actively looking to deceive with AI-generated content will look for ways to circumvent these safeguards, so we need to continue to look for ways to stay one step ahead," Clegg said.