Meta to change the 'Made with AI' label to 'AI info' for AI-edited photos

Meta recently faced criticism after introducing a 'Made with AI' label on tagged photos in May with photographers expressing dissatisfaction, noting that the label was being applied to real photos that had undergone basic editing rather than being solely AI-generated. Responding to user feedback and acknowledging widespread confusion regarding the level of AI involvement in photo creation, Meta has decided to replace the label with 'AI info' across all its apps.

In a blog post addressing the issue, the company admitted that the original tag failed to clearly communicate to users that photos with the label may not have been entirely AI-generated but could have involved AI-powered editing tools. The company explained that its previous labels did not always align with user expectations and lacked sufficient context. For instance, even minor adjustments made using AI-powered retouching tools were labelled as 'Made with AI,' which contributed to the misunderstanding.

Meta's decision to switch to 'AI info' aims to provide clearer and more accurate information about the role of AI in the creation of tagged photos, aiming to better meet user expectations and alleviate confusion within the community and industry standards.

The company is maintaining its current technology for detecting the use of AI in photos and applying appropriate labels, despite changing the label itself. The company continues to rely on technical metadata standards like C2PA and IPTC, which provide information regarding the use of AI tools.

For example, if photographers utilise tools such as Adobe's Generative AI fill to remove objects from photos, their images may still receive the new 'AI info' label. However, Meta aims to clarify that this label does not imply that the entire image was created solely by AI.

Moving forward, Meta and other social networks will need to establish guidelines that are equitable for photographers who have integrated AI-enhanced editing tools into their workflows.