Many e-commerce businesses today use AI to create images for blog posts, campaigns, and product descriptions. According to new regulations from the EU, there is now a requirement that content created by AI which risks misleading the recipient must be marked with a label.
For a label to be approved, it must directly and without room for interpretation inform the recipient that the material is computer-generated. This can be done through a clearly visible text in connection with the image, for example, a short and concise text stamp on the edge of the image.
The labeling should be clear enough to deter deception, but it does not need to cover the entire image and thereby ruin the visual experience.
Another approved approach is embedded digital watermarks which allow programs to automatically identify the files.
Only misleading material is affected
The responsibility for complying with the rules lies with the creators of the systems and the professional users.
E-commerce businesses that publish realistic campaign images must therefore ensure that customers are informed. At the same time, certain material is exempt to avoid creating label fatigue in the recipient, where the warnings eventually lose their effect.
The EU has chosen to focus the legislation on content that can actually deceive the eye in a commercial or news context.
It is difficult to know what to trust, what is real and what is not real. It is both a democratic problem and a consumer problem," says Katja de Vries, Associate Professor of Public Law at Uppsala University, to Sveriges Radio.
Risk of fines in e-commerce
If a company violates the AI regulation, sanction fees may be imposed. The penalty can amount to 3 percent of total turnover, or 15 million euros.
However, lower requirements are placed on how prominent the labeling must be for content with an obvious artistic or satirical purpose.