AI policy

The short version

• Artists declare. Work here carries the artist's own word for how much AI was in its making: AI-free, AI assisted or AI generated. We never run detectors.
• We never train on your photos. GallerySpark never trains anything on artists' photos — not for our own tools and not for anyone else's.
• It is about the art itself. AI in the business around the art — emails, invoices and promotion — never counts, and assistive technology never needs declaring.

The three levels

One vocabulary, used the same way everywhere on GallerySpark.

AI-free

No AI in the making of the work itself.

AI assisted

The artist made the work and AI touched part of the process. That includes AI features inside ordinary editing software: generative fill on a corner, cleanup and upscaling are all assisted. What never counts is AI in the business around the art — the emails, invoices and promotion. The declaration is about the art itself, never the paperwork.

AI generated

AI produced the work's expressive content and the artist's part was directing and selecting. The line is authorship — CARFAC states it as the author's "skill and judgment" versus a "purely mechanical exercise".

Assistive technology is carved out entirely. Voice-to-text and tools like it are not AI use in a work and never need declaring.

Where the levels live: a gallery declares the level it expects to include, each work declares its own and a call declares what it accepts. Those declaration choices are being added to Gallery settings, the upload screens and call setup — this page is the reference they point to.

Declared, never detected

There is no AI detection on GallerySpark, ever. Detectors are unreliable and a false accusation is worse than a wrong declaration. The policy runs on honesty, stated plainly: artists declare, and the declaration is what visitors see.

Where this comes from

• The Canada Council's guidance (December 2025) asks artists to disclose AI use inside funding applications — narrative disclosure, no set levels and no public label.
• The federal Heritage Committee's 2026 report recommends standardized labelling, especially for fully synthetic content.
• CARFAC's survey found 82% of artists very or extremely concerned about non-consensual training on their work. That concern is why the no-training promise sits at the top of this page.

And this site itself

By these levels, GallerySpark is an AI assisted work: a human directed it and made the decisions, and AI did a great deal of the making. The policy applies to us too, so we declare it here.

We used AI to build this site, because it is about time AI did something for artists and not the other way around.