What can I do to prevent my work from being used as training data for an AI system?
As a creator, you probably cannot prevent your work from being "scrapped" from the Internet, and used for AI training, if this is done by non-profit scientific research organizations and cultural heritage institutions. The Copyright Act does allow you to "opt-out" of commercial text and data mining. However, this must be done in a way that is readable by the TDM software.
There are several ways to do this. First, you can likely do this by using the so-called "robots.txt protocol" on your Web site. LAION, the database behind Stable Diffusion AI, has indicated that its crawler respects the robots.txt protocol. However, the downside of this is that it probably limits your visibility online, and it cannot be used on platforms, for example, but only on your own web page. Online you should be able to find more information on exactly how to do this.
Additionally, you could use fixed settings on platforms where a lot of visual work is shared. For example, ArtStation and DeviantArt have now developed a 'tag' called NoAI or NoImageAI tags that you can select when uploading your work on these platforms. With this, you indicate that you are exempting this particular work from being used for TDM. This is integrated into the HTML, and is readable by scraping software.
Other ways to opt-out are still under development. People from across Europe are also looking for a universally accepted way to shape the opt-out. Pictoright and its partners are monitoring these developments.