Meta Revises AI Content Labeling Approach on Facebook and Instagram

Meta Hides 'AI Info' Labels for Edited Content on Facebook, Instagram

Meta is reducing the visibility of "AI Info" labels on Facebook, Instagram, and Threads. Starting next week, for content modified or edited using AI tools, the label will be found in the post's menu, the company announced in an updated blog post on Thursday. This means the labels won't be quite as visible as before. The revised policy only applies to AI-edited content, the company says. For content generated using an AI tool, Meta will continue to display the "AI Info" label under the username and share whether media is labeled "because of industry-shared signals or because someone self-disclosed." Once the update rolls out, to check whether the content was altered using AI, users will have to tap the menu button at the top right of a post and pick "AI Info." The change is intended to "better reflect the extent of AI used in content," says Monika Bickert, Meta's Vice President of Content Policy. "We've continued to work with companies across the industry to improve our labeling process so that labels on our platforms are more in line with peoples' expectations." Meta introduced the "AI Info" tag in July after its predecessor, "Made with AI," faced harsh criticism for being confusing and inaccurate. The company now determines the labeling based on "a signal added by the tool used to modify the content. This signal can indicate a wide range of Al use, from photo retouching to generating entirely new content," according to the AI Info dialog box. However, Meta warns their labeling still isn't always accurate because not all Al media has the metadata they need to flag it.

Meta is making its AI info label less visible on content edited or modified by AI tools

Meta is changing the way it labels content that has been edited or modified by AI tools on Instagram, Facebook, and Threads. For this type of content, Meta is moving the "AI info" label to the post's menu. In the past, the label would appear directly under the user's name. The company says the label will still appear under content that it has detected was generated by an AI tool. This means that although the label is being hidden for content that was changed or edited by AI tools, it will still be prominently displayed under content that was fully generated by an AI prompt. For content that was generated by AI, Meta will "share whether the content is labeled because of industry-shared signals or because someone self-disclosed," the company says. Meta says the change, which is rolling out next week, will "better reflect the extent of AI used in content" on its platforms. By making the AI info label harder to find, it might be easier for users to be deceived by content that was edited with AI, especially as editing tools become more and more advanced. Given that generative AI is a relatively newer technology, this isn't the first time that Meta has changed how it labels such content on its platforms. In July, the company changed its AI label from "Made with AI" to "AI info" after Meta complaints from photographers who said the label was being added to real photos. At the time of the change, Meta said the original wording of the label wasn't clear enough for users to understand that certain images with the tag were not necessarily created with AI, but may have been edited using AI-powered tools.

Meta hides warning labels for AI-edited images

You'll have to go deep into a post's information menu to see if it has been edited using AI tools. Starting next week, Meta will no longer put an easy-to-see label on Facebook images that were edited using AI tools, and it will make it much harder to determine if they appear in their original state or had been doctored. To be clear, the company will still add a note to AI-edited images, but you'll have to tap on the three-dot menu at the upper right corner of a Facebook post and then scroll down to find "AI Info" among the many other options. Only then will you see the note saying that the content in the post may have been modified with AI. Images generated using AI tools, however, will still be marked with an "AI Info" label that can be seen right on the post. Clicking on it will show a note that will say whether it's been labeled because of industry-shared signals or because somebody self-disclosed that it was an AI-generated image. Meta started applying AI-generated content labels to a broader range of videos, audio and images earlier this year. But after widespread complaints from photographers that the company was flagging even non-AI-generated content by mistake, Meta changed the "Made with AI" label wording into "AI Info" by July. The social network said it worked with companies across the industry to improve its labeling process and that it's making these changes to "better reflect the extent of AI used in content." Still, doctored images are being widely used these days to spread misinformation, and this development could make it trickier to identify false news, which typically pop up more during election season.

Instagram is Changing How It Labels AI Content, Again

Earlier this year, Meta -- parent company of Facebook and Instagram -- said it would be adding a notification to images uploaded to its platforms that were made with AI. That rollout went poorly, and now the company is making another update to how it displays that notification, it's second in three months. Initially, Meta made these changes in response to recommendations from its Oversight Board, which said that giving users "transparency and additional context is now the better way to address manipulated media and avoid the risk of unnecessarily restricting freedom of speech." In response to that recommendation, Meta went with a very broad brush, labeling anything that came with the Content Authenticity Initiative's AI tags as "Made with AI" even if it was only a small adjustment to the image. As PetaPixel reported during the initial rollout in May, the "Made with AI" label created more confusion than it cleared up and resulted in widespread outcry from photographers and content creators. In July, Meta shifted its tag from "Made with AI" to "AI Info," but didn't adjust what types of images would get this tag. In this way, the approach would still punish photographers who use Adobe Photoshop equal to those who use full generative AI programs, like Stable Diffusion. Additionally, the tags Meta is reading -- the ones from the CAI -- aren't included on all AI-generated images, letting some slip past Meta's system. Additionally, Meta isn't part of the CAI, meaning it's not actively working with the companies that are trying to clear up AI confusion -- it's just using their work. At least part of that approach is changing. Yesterday, Meta announced that it would be adjusting the "AI info" label to appear less penalizing to photographers who use an AI feature in Lightroom or Photoshop to adjust a real photo. "Soon we're rolling out a change to the 'AI info' labels on our platforms so they better reflect the extent of AI used in content. Our intent has always been to help people know when they see content that was made with AI, and we've continued to work with companies across the industry to improve our labeling process so that labels on our platforms are more in line with peoples' expectations," Meta writes. "For content that we detect was only modified or edited by AI tools, we are moving the 'AI info' label to the post's menu. We will still display the 'AI info' label for content we detect was generated by an AI tool and share whether the content is labeled because of industry-shared signals or because someone self-disclosed. These changes will begin next week."

Facebook and Instagram are making AI labels less prominent on edited content

The company says the changes are being introduced to "better reflect the extent of AI used" across images and videos on the platforms. This label was introduced in July after Meta's previous "Made with AI" label was criticized by creators and photographers for incorrectly tagging real photos they had taken. "We will still display the 'AI info' label for content we detect was generated by an AI tool and share whether the content is labeled because of industry-shared signals or because someone self-disclosed," Meta said in the update, adding that the changes will start rolling out next week.