<aside>
<img src="/icons/alien-pixel_blue.svg" alt="/icons/alien-pixel_blue.svg" width="40px" /> How to take down images from generative AI model datasets
</aside>
<aside>
⚠️
Goodbye Meta AI
Posting a “Goodbye Meta AI” message on Facebook, Instagram, or WhatsApp does not prevent Meta from training on user content. Making your content private will prevent future AI training but does not reverse previous training.
</aside>
ArtStation
- ArtStation allow users to label projects as
NoAI to indicate that they are not authorized for third-party AI training – how to label
Canva
DeviantArt
- DeviantArt’s DreamUp image generator is not trained on user content, but allows users to generate images in the style of other DeviantArt users’ work, while giving them credit
- To prevent users from including your username and other pseudonyms in DreamUp prompts, opt out here
- User images are automatically marked with
NoAI labels, which indicate that they are not authorized for third-party AI training – how to edit NoAI labels
Figma
- Starter and Pro teams are opted in by default to Figma content training, while Organization and Enterprise teams are automatically opted out by default – how to opt out
Tumblr
- In February 2024, 404 Media reported that Automattic, the parent company of Tumblr, was preparing to strike deals with Midjourney and OpenAI for user content.
- That same month, Automattic published a blog post which stated that it is sharing public content hosted on Tumblr with select AI companies but will respect opt-outs – how to opt out
WordPress
- In February 2024, 404 Media reported that Automattic, the parent company of WordPress, was preparing to strike deals with Midjourney and OpenAI for user content
- That same month, Automattic published a blog post which stated that it is sharing public content hosted on Tumblr with select AI companies but will respect opt-outs – opt out of content sharing