Pinterest’s doubling down on Generative AI threatens to alienate tabletop artists and publishers

Pinterest has updated its global Terms of Service and Privacy Policy to give itself explicit permission to use uploaded Pins, board collections, and user descriptions to train proprietary machine learning and generative AI models. The revised terms apply retrospectively across all historical uploads, regardless of when an image was posted, creating an immediate copyright clash with tabletop roleplaying game publishers and independent illustrators who reject generative tooling.

While Pinterest has introduced a user-facing toggle to opt out of training its Canvas generative engine, the mechanism offers weak protection. The platform’s baseline licence treats uploads as perpetual and sublicensable, meaning human-created artwork remains vulnerable to automated ingestion the moment another account saves or re-pins it.

The Tabletop Industry’s Zero-Tolerance Stance on Machine Learning

Many in the tabletop roleplaying game sector maintain an exceptionally strict boundary against synthetic media. Players and indie studios actively boycott products featuring automated writing or generative imagery, treating human creative labour as non-negotiable.

This sensitivity is so pronounced that even automated layout utility tools cause uproar. Earlier this year, Adobe InDesign rolled out an unannounced tool that quietly inserted automated alt text into document assets. When fans spotted synthetic image descriptions inside Chaosium’s Time Without Tide quickstart on BackerKit and Evil Hat Productions’ digital edition of Umdaar: Rebel Broadcast Edition, both publishers faced immediate scrutiny. Both studios moved quickly to strip out the tool’s output and reassure backers of their stance. As Evil Hat’s Fred Hicks and Chaosium’s Brian Holland made clear, the community expects zero machine-learning involvement in roleplaying games – even down to accessibility labels.

Given this environment, indie illustrators, cartographers, and writers will not tolerate having their work fed into image generation engines. Many rely on Pinterest as an informal visual portfolio, only to discover that years of back-catalogue illustrations have been converted into training data by default.

Why the Canvas Opt-Out Fails Creators

Pinterest added an opt-out control located under Settings > Privacy and data titled “Use your data to train Pinterest Canvas”. However, platform architecture severely undermines this protection:

• The Re-Pin Problem: If you upload an original battlemap or character portrait and toggle the setting off, Pinterest will not train directly from your specific account. The moment another user saves that Pin to their public board, their account settings govern the image. If they have not toggled off the Canvas setting, the asset enters the training pipeline anyway.
• Persistent Sublicensing: Pinterest’s Terms of Service demand a non-exclusive, royalty-free, transferable, and sublicensable licence to host and distribute content. Deleting your original Pin or closing your account does not remove copies that other users have saved to their boards.
• Retrospective Default: Accounts default to opted-in. Millions of historical TTRPG sketches and commissions uploaded over the past decade are already exposed unless creators manually log in to adjust their preferences.

How Website Owners and Publishers Can Disable Pinterest

Publishers, artists, and independent blogs can block Pinterest’s automated ingestion tools at the domain level. Adding code directly to your site prevents browser extensions, user bookmarklets, and native crawler scrapers from harvesting original art.

The nopin meta tag actively disables the Pinterest save extension, while adding a custom description attribute forces a pop-up alert explaining why pinning is barred when someone tries to grab the image. Pairing this with a Disallow rule in robots.txt prevents Pinterestbot from scraping Open Graph preview images (og:image) directly from article URLs.

While this does not stop someone from manually saving an image to their hard drive and uploading it as a raw file, it closes off browser-level pinning and automated syndication.

A Risky Move for Visual Discovery

Pinterest built its market identity as a digital haven for visual curation, human discovery, and mood-boarding. Unlike algorithmic video feeds, its core utility relied on real people categorising interior ideas, craft projects, miniature paint schemes, and hand-drawn maps.

Transforming that human library into a training engine threatens the platform’s foundational value. Flooding search results with unlabelled synthetic imagery alienates the artists whose work populated the ecosystem in the first place. For indie RPG creators, the update makes Pinterest look less like an inspirational discovery tool and more like an extraction pipeline. Studios and freelancers will now have to weigh whether occasional referral traffic justifies surrendering their intellectual property to a model.

Image credit: Arseny Togulev on Unsplash.

Quick Links

Pinterest: Terms of Service.