Kickstarter pulls plug on AI porn project Unstable Diffusion

If you remove the safety filters, uncensored AI can pull in cash at an astonishing speed.

Kickstarter, however, doesn't want to touch that money. On December 21, the crowdfunding giant abruptly shut down a campaign for Unstable Diffusion, an initiative focused on training generative AI models specifically for NSFW and adult content. The group had asked for $25,000 and easily crushed that goal, pulling in over $56,000 from 867 backers before the plug was pulled.

Because Kickstarter runs on an all-or-nothing model, the suspension came before the campaign concluded. That means every single dollar will be refunded, and the Unstable Diffusion team walks away empty-handed.

Artists strike back over copyright and consent

Kickstarter CEO Everette Taylor took to the company blog to defend the decision, stating that the platform will always stand on the side of human creators and creative work. While the company's broader policy on AI remains vague, Taylor noted that Kickstarter is actively evaluating how projects handle copyrighted training data and whether they pose safety risks to communities.

Backlash from the creative community forced Kickstarter's hand. High-profile illustrators Karla Ortiz and Sarah Andersen openly called out the platform on Twitter, arguing that backing the project amounted to funding rampant intellectual property theft and tools that enable nonconsensual imagery.

Artists were particularly alarmed by Unstable Diffusion's own pitch. The campaign bragged about using the funds to train an expanded model on 75 million images, including 25 million pulled directly from art hubs like ArtStation, DeviantArt, and Behance.

Much of the pushback centered around artists like Greg Rutkowski, a fantasy illustrator whose signature style became one of the most scraped prompts on Stable Diffusion without his consent. For creators, watching an adult AI team monetize their uncredited work was simply a step too far.

Banned from Kickstarter, funded on Discord

Losing the Kickstarter campaign barely slowed the project down. Arman Chaudhry, CEO of Unstable Diffusion, immediately mobilized the group's Discord community of over 97,000 members and routed them to a self-hosted checkout using Stripe.

Within hours, Chaudhry claimed the project had already raised over $15,000 directly. He framed the Kickstarter ban as a cowardly capitulation to an organized anti-AI lobby, claiming traditional artists had already raised $150,000 on GoFundMe to push for restrictive government regulations.

Technically, base models like Stable Diffusion only feature around 2.9% NSFW data, making unassisted adult imagery difficult to generate. Unstable Diffusion relies on volunteer communities to scrape and label specialized datasets to fine-tune the weights, promising improved anatomy and broader styles.

Yet deep ethical traps remain. A vast amount of web-scraped adult imagery lacks clear consent from the subjects involved. Without strict safeguards, those same models can be trivially weaponized to create nonconsensual deepfake pornography, an area where legal protections remain virtually nonexistent across most jurisdictions.

Why it matters

Kickstarter shutting down Unstable Diffusion shows that mainstream tech platforms are no longer willing to host the messy wild west of generative AI without blowback.

For traditional artists, this is an undeniable tactical win. Public pressure worked, proving that organized pushback can cut off funding rails from commercial operations built on scraped art. But for users looking for open-ended, uncensored tools, the writing is on the wall: explicit and frontier-pushing image generators will increasingly be relegated to decentralized servers, direct crypto rails, and private web checkouts.

Kickstarter's move was smart for brand preservation, but fundamentally cowardly in its execution. By adopting intentionally vague criteria, the platform dodged setting a hard precedent on AI art while appeasing its core creative base.

We think kicking these projects off mainstream platforms solves very little in practice. Unstable Diffusion recovered thousands in direct funding within days. Deplatforming doesn't stop uncensored models from existing; it simply pushes development into shadowy corners where accountability, safety filters, and consent verification are even harder to enforce.