Instagram, Threads, and Facebook will start labeling AI-generated content

Meta is planning to make it easier to tell if content was made with AI on Threads, Instagram, and Facebook. To accomplish this, the company says it will start labeling such content “in the coming months.”

Today, Meta announced it is working on building the capability to automatically detect signals that tell if a piece of content was made with AI. Once ready, this capability will allow the company to put labels on content posted to Facebook, Instagram, and Threads. Additionally, the labels will come in all languages supported by each app.

When the capability is rolled out, it will affect images posted on Meta’s platforms. It will work by identifying industry standards used for identifying AI images, such as invisible watermarks, embedded metadata, visible markers, and so on. If an image doesn’t have these signals, Meta warns that those images may not be spotted.

This capability won’t be limited to just identifying images made with Meta’s AI tools. The company is aiming to be able to identify AI-generated images created with AI tools from other companies — like Google, OpenAI, and Adobe — as well.

It’s not just images that Meta is focused on; the company also wants to label AI-generated videos and audio. However, as Meta points out, these signals haven’t been included in AI tools that produce video and audio to the same extent. Until they do, Meta says in the meantime, it will introduce “a feature for people to disclose when they share AI-generated video or audio so we can add a label to it.”

Labeling videos or audio as generated by AI will not be optional. Meta says it will require users to use this disclosure and label tool when posting “a photorealistic video or realistic-sounding audio that was digitally created or altered.” The company further adds that it may punish users if they discover AI-generated videos or audio that weren’t labeled.