The tech multinational organization is responding to concerns raised by photographers about potential copyright infringement by modifying its identification tags on AI-generated imagery. This alteration is part of a broader commitment to respect the professional work of photographers and ensure the company’s compliance with copyright laws.
Previously, restrictions resulted in slightly revised images being incorrectly tagged as “Made with AI”. This led to confusion and stress for users, who were mislead into believing that these modifications were a product of artificial intelligence, not the effect of restrictions in place. The organization has taken note of these issues and is now actively rectifying this problem.
The initial “Made with AI” labels were rolled out earlier this year to denote photos produced by generative AI. However, even minor adjustments via editing tools, like Adobe’s generative fill tool, could trigger these AI labels.
Refining AI image labeling protocols
Users felt the tool was overly sensitive and argued that minor alterations should not result in a “Made with AI” designation.
In light of this feedback, the organization has significantly updated its system to adjust the sensitivity levels of its AI detection. Now, only major changes made by AI will trigger the “Made with AI” label, thereby eliminating the chance of wrongfully labelling minimally edited images. This change offers a more accurate representation of human vs AI contribution to the creative process.
To clear up previous confusion, the company is also planning to rename the “Made with AI” tag to “AI info”. This rebranding effort aims to provide users with a clearer understanding of the AI processes involved in image alterations.
This amendment reflects an industry-wide initiative to improve the labeling process to better align the technology used with the labels applied. Ideally, these modifications will not just enhance clarity and precision, but also significantly improve the overall labeling accuracy.
However, while the newly designated “AI info” labels will indicate AI involvements, they won’t specify which AI tools were used to modify the image. Furthermore, a contextual menu highlighting a general overview of generative AI will still be available to users.
