Meta has transitioned its photo labeling system from “Made with AI” to a new “AI Info” tag across its platforms. This decision comes after photographers and users expressed that the initial label created confusion.
The “Made with AI” label was applied to authentic photos with minor edits, leading to misconceptions regarding the scope of AI involvement. Photographers found this misleading, prompting Meta to reconsider its labeling approach. The former tag failed to differentiate between fully AI-generated images and those with slight AI-assisted modifications.
Technical Metadata and Detection Mechanisms
Although the label has changed, the technology identifying AI usage remains the same. Meta utilizes metadata standards like C2PA and IPTC, which track AI tool usage in photos. Images edited through Adobe’s Generative AI Fill to remove objects will now carry the “AI Info” label instead of the previous tag.
Meta aims to provide more accurate context with the “AI Info” label, addressing user expectations more effectively. The company plans to collaborate with other industry entities to refine these guidelines further. According to Meta’s representative, Kate McLaughlin, this new tag will cover content both created and edited with AI, ensuring users have a clearer understanding of how their content has been modified.
Addressing Broader Challenges
The update does not solve the issue of completely AI-created images slipping through undetected, nor does it detail the level of AI editing applied. Meta, along with other social networks, is developing guidelines that maintain transparency without negatively affecting photographers who use standard AI-based editing tools.
Meta launched the “Made with AI” labels earlier this year in response to feedback from the Oversight Board concerning its “manipulated media” policy. The company noted that content with minor AI adjustments, signaled by industry-standard markers, would trigger the “Made with AI” tag. The contextual menu accessed by clicking the “AI Info” badge will remain the same, offering a general overview of generative AI and indicating that Meta might flag content with AI signals detectable by their systems.
Last Updated on November 7, 2024 3:43 pm CET