Are You Labeling Your AI Images?
Have you noticed a small icon (cr) appearing on LinkedIn images and elsewhere? We're currently navigating an "AI content labeling jungle," and for businesses, it can be challenging to find the right path.
DO AI IMAGES NEED TO BE LABELED?
SCENES Code of Conduct is our attempt to help clients address this issue. In short, we see four possible approaches:
1) ICON + INFORMATION
The C2PA standard, supported by companies like OpenAI, BBC, Google, Adobe, Microsoft—and SCENES (a member since April 2024)—provides a solution.
ADVANTAGE: Highly visible with the cr-icon directly on the image, offering easy access to creation and modification details.
DISADVANTAGE: Visually intrusive.
2) WATERMARK + INFORMATION
Google DeepMind’s SynthID applies an invisible digital watermark that works across most media types.
ADVANTAGE: Invisible on the image but still provides traceable creation details.
DISADVANTAGE: No visible marker; users must actively search for info (e.g., by right-clicking the file).
3) TEXT
Similar to how ad agencies used to print their names discreetly in the corner of ads.
ADVANTAGE: Subtle visibility, suitable for non-digital formats like print and billboards.
DISADVANTAGE: Easy to overlook—banner blindness and no additional information.
4) NOTHING!
There’s no legal requirement in Denmark to label AI-generated images.
ADVANTAGE: Users aren’t distracted by “warnings” or creation details. When was the last time you saw an ad explaining its production process?
DISADVANTAGE: Doesn’t help demystify AI or reassure those concerned.
Make a Choice!
When SCENES creates an image for a client, we transfer the rights to them—something we’ve done for over 20 years. What’s new is that companies must now decide how, or if, their content should be labeled for their audience.
Personally, I’m comfortable with all four approaches, as long as the company makes an informed decision.
Michael Rying
