In separate discussions with US media, Adobe has shed some light on its confusing privacy stance. Adobe was criticized after some users of the company’s cloud services discovered a switch that was only supposed to prohibit the use of their content for computer-generated media by opting out.
As a spokesman for the company now said petapixel, the assumption is incorrect that images and videos in the Adobe cloud are used as training data for new content. “For Generative AI, Adobe does not use data stored in the Creative Cloud by users to train its experimental Generative AI capabilities.”
The term “Generative AI” has become common in the USA for systems that generate new content from templates. Many artists and media professionals had already complained about systems like Dall-E, where it is not always clear what content they were trained on. It was therefore also feared that Adobe, which probably has one of the largest treasure troves of image and video data, could use it to train its models.
User data has never been used for Generative AI
At Bloomberg became even clearer independently of the report by Petapixel Adobe product chief Scott Belsky: “We have never used anything from our memories to train a model for generative AI.” That had “not even” happened. From the context it becomes clear that the “storage” must mean the Creative Cloud, because Adobe has to work on its admittedly existing AI models on the basis of some kind of storage.
According to Belsky, the criticism of the supposedly new data protection options was a “wake-up call”. If Adobe allows users to release their data for Generative AI in the future, this must be clearly marked and it must also be explained how the data is used. Belsky explicitly mentioned the term “opt-in”, so far the settings in an Adobe account gave the impression that the use of AI had to be explicitly contradicted by “opt-out”.
In the statements, the Adobe representatives only go into Generative AI – but not whether user content is also used for other forms of machine learning. What is behind the controversial switch remains unclear. A spokesman for Adobe said that the function is used, among other things, for object recognition in Lightroom, for example if you tag a photo with the keywords “dog” or “cat”. In general, as Scott Belsky told Bloomberg in return, his company is working on new data protection settings, which should then be explained to users in a more understandable way before any decisions are made.
(keep)
