(Credit: Koshiro K/Shutterstock.)
Meta will begin adding labels on AI-generated content "in the coming months" in an effort to help users quickly understand what's real and what's not across Facebook, Instagram, and Threads, the company announced Tuesday.
"Our users have told us they appreciate transparency around this new technology," Meta President of Global Affairs Nick Clegg wrote in a post. "It’s important that we help people know when photorealistic content they’re seeing has been created using AI."
Meta already labels content generated using its MetaAI feature as AI-generated. Now, it's also prepping a tool to determine whether any images on its social media apps were made with AI.
Only images that already include hidden watermarks and metadata identifying their AI-generated status can be captured and labeled. Meta says it will soon be able to label images made with AI tools from Google, OpenAI, Microsoft, Adobe, Midjourney, and Shutterstock—but only after these platforms add self-labeling metadata. This means that any content created before these firms add metadata won't be detectable under Meta's method.
Another big caveat is that Meta says it hasn't yet cracked how to determine if audio and video are AI-generated at scale. "We can’t yet detect those signals and label this content from other companies," Clegg said of audio and video content.
In the meantime, Meta will ask users to disclose whether their content is AI-generated.
"We’ll require people to use this disclosure and label tool when they post organic content with a photorealistic video or realistic-sounding audio that was digitally created or altered, and we may apply penalties if they fail to do so," Clegg said.
Meta's efforts to increase transparency around AI-generated media is a step forward for media literacy overall, but its limited current scope of detection won't be much of a solution for combatting intentionally deceptive AI-generated posts created with tools that don't have "invisible markers."
PCMag has reached out to Meta for additional comment on the steps it's taking, if any, to label AI-generated content that doesn't contain detectable markers.