Tech Desk 
Sep 13, 2024
Meta is working with ways to make artificial intelligence (AI) generated content more transparent and accurate. The company recently announced changes to its AI labelling policy as a way to give users more granular information about the content with which they engage.
Currently, Meta uses an "AI Info" tag for all content that has anything to do with AI, whether the image was lightly edited or completely created by AI. However, this has largely been seen as a misstep as some real photos have still been tagged.
To counteract this, Meta would introduce labelling. Alongside the "AI Info" tag, a menu in the top right of pictures and videos that have been edited by the use of AI, as opposed to directly underneath the user's name, is the next development planned.
Users can tap on this menu to see whether AI info is available to see and to read what may have been adjusted.
Read more: WhatsApp introduces Meta Verified, customised messages in India
While still showing the "AI info" label for content detected to be AI-generated, Meta would take off these tags from real images that have been manipulated by AI tools. This change shall bring a better representation of where and how much AI is used across images and videos on the platforms.
Meta also points toward "industry-shared signals" that help identify AI-generated content. These include systems such as Adobe's C2PA-supported Content Credentials metadata and Google's SynthID digital watermarks.
While the new labelling system holds better information, taking tags from real images that have been manipulated would make things further from the ability to let users avoid being misled.
