The application · 11 min read

AI and Art Open Calls: What the Rules Say and How to Stay Safe

Most calls are silent about AI, a few ban it, and public funders put the responsibility on you. How to read the rules, disclose clearly, and protect yourself from false suspicion.

Published · CallOpen research team

In short

Most open calls say nothing about AI: of 1,608 open calls in our catalogue, 95% did not mention it, 61 banned AI-generated work, 8 required disclosure and 5 explicitly welcomed it (28 October 2026). Bans cluster in competitions, especially photography; World Press Photo 2026 bans generative AI in every category and checks finalists’ RAW files. Public funders such as Canada Council and Arts Council England allow AI tools but make you responsible for what you submit. Silence is not permission: ask.

  • —If a call is silent, that is uncertainty, not permission. A short question to the organiser removes it.
  • —Disclosing is safer than hiding. A good disclosure says what, which tool, at what stage, and what you did yourself.
  • —Detectors are unreliable. Your best defence against false suspicion is process: originals, sketches, stages, layers.
  • —Calls that welcome AI are rare. If AI is your medium, look for media-art festivals and art-and-technology programmes.
Calls silent about AI
95%
1,534 of 1,608 open calls, 28 October 2026.
Calls banning AI-generated work
61
Competitions 40, exhibitions 13, fellowships 4, grants 2, other 2.
Calls requiring disclosure
8
And 5 explicitly accepting AI work.
EU labelling duties apply from
2 August 2026
Article 50 of the AI Act, for synthetic content.

How we counted: AI rules were read from the organiser’s terms for every open call in the CallOpen catalogue on 28 October 2026. Legal points are general information as of October 2026, not legal advice; check official texts before relying on dates in a contract.

What the rules say

Bans concentrate where a single work is judged. Competitions account for 40 of the 61 bans, exhibitions for 13. For a competition, “who made this” is the core question. Grants and fellowships judge the artist and the project, so their question is whether the application is truthful, not who made one image.

Photography has the strictest rules. World Press Photo’s 2026 rules ban generative AI in every category, including generative fill and AI upscaling, and finalists must provide their original files. Sony World Photography Awards 2026 requires disclosure of AI manipulation and bans excessive use. The line is moving from “digital versus analogue” to “editing what was captured versus generating what was not”.

Public funders take another route. Canada Council and Arts Council England say they do not use AI to assess applications. They do not forbid applicants from using AI tools, but the applicant is responsible for the accuracy and authenticity of everything submitted. That is a good rule to adopt with any funder: you can use the tool, you cannot hand it the responsibility.

Five levels of AI use, five levels of risk

“AI” covers very different actions, and rules and juries treat them differently. Be clear with yourself about which level you are on (see the table below). Spell-checking or translating your own draft is a different thing from generating an image from a prompt.

If AI is your medium, look for calls that welcome it (only 5 did on 28 October 2026), media-art festivals and art-and-technology programmes. There, describing your method is not a confession; it is part of the concept.

What the law says in 2026

General information, not legal advice. Authorship: in the US, human authorship is current law. The appeals court in Thaler v. Perlmutter (18 March 2025) confirmed that a work made without a human author is not protected by copyright, and on 2 March 2026 the Supreme Court declined to hear the case. The US Copyright Office’s 2025 report says AI-assisted work can be protected to the extent it expresses a human’s creative contribution; a prompt alone generally is not enough.

Training: in the UK, the High Court on 4 November 2025 rejected the main claims of Getty Images against Stability AI. In the US, artists’ case Andersen v. Stability AI, Midjourney, DeviantArt and Runway continues in a narrower form, with summary-judgment motions set for 17 February 2027. There is no ruling on the merits yet.

Labelling: in the EU, transparency duties for synthetic content under Article 50 of the AI Act apply from 2 August 2026, with a delay to 2 December 2026 for machine-readable marking by generative systems already on the market, according to legal reviews.

The practical point: organisers ban AI not because the law requires it, but because their selection and their audience’s trust rest on human authorship.

How to decide and how to disclose

Search the rules, FAQ and form for “AI”, “artificial intelligence”, “generative”, “machine learning”, “IA”, “KI”. If there is a ban, find out exactly what is banned: generation, editing, or texts. If your work sits on the line (AI noise reduction in a photo, for example), ask the organiser in writing. If disclosure is required, write a short, concrete one. If the call is silent: routine help with texts usually needs no disclosure, as long as the facts are yours; AI inside the work itself is worth asking about or disclosing voluntarily in the work description.

A good disclosure says what, with which tool, at what stage, and what you did. For a painting: “The composition is based on sketches I made by hand. While choosing the palette I used a generative model to produce colour variations; the final work is oil on canvas and contains no generated images.” For a text: “I wrote this text; a machine translator was used for a first English draft, which I edited.” For AI as medium: “The video is built from images generated by a model I trained on my own photo archive (2015–2024). Editing, sound and script are mine.”

Avoid vague phrases (“with elements of AI”), minimising, and legal-sounding disclaimers that say nothing.

Detectors are unreliable. Process files are not

Tools that claim to detect AI images and texts produce false positives on human work and miss generated work. The C2PA standard (Content Credentials) can attach provenance and editing history to a file, but it confirms a chain of processing, not that a human made the image: a missing label proves nothing, and labels are easily lost on export.

Your strongest defence against false suspicion is the record of your process: original files, sketches, photographs of the work at different stages, process videos, layered files. World Press Photo checks RAW files precisely because they are the hardest evidence to fake. Keep them for every work you submit.

Texts carry a separate risk. Models confidently invent exhibition titles, dates and quotes. Every line of a CV or proposal must be checked by you, and a vague text is easy to spot with the name-swap test long before any detector.

Read the rights clause too. If a call asks for a licence to use submitted material for “machine learning” or “technology development”, that is a separate grant of rights, and you may refuse it.

Levels of AI use and risk in open calls
LevelExamplesRisk in open calls
RoutineSpelling, translating a draft, shortening, schedules, budget tablesMinimal, if the facts are yours and checked
PreparationFinding references, generating composition variants as sketches you then make yourselfLow; disclose if the call asks
Editing images of the workAI noise reduction, upscaling, object removalMedium to high: often banned in photography
Part of the workGenerated elements in collage, video or installation; models as a mediumHigh where banned; fine where AI is allowed or is the theme
Fully generatedAn image from a prompt without substantial further workBanned in many competitions; copyright protection doubtful

Template

AI disclosure template

Fill in only what applies. Keep it short and specific; put it where the call asks, or in the work description if the call is silent.

AI DISCLOSURE — [work title / application]

What I made myself: ____
Where AI was used (stage): ____
Tool(s): ____
What the tool did, exactly: ____
What I did with the output: ____
What contains no generated material: ____

EXAMPLES
Painting: "The composition is based on sketches I made by hand. While choosing the palette I used a generative model to produce colour variations; the final work is oil on canvas and contains no generated images."
Text: "I wrote this text; a machine translator was used for a first English draft, which I edited."
AI as medium: "The video is built from images generated by a model I trained on my own photo archive (2015–2024). Editing, sound and script are mine."

BEFORE SUBMITTING
[ ] Searched the rules for AI / artificial intelligence / generative / machine learning / IA / KI
[ ] If banned: I know exactly what is banned; borderline uses asked about in writing
[ ] Every fact in my texts checked by me
[ ] Process files kept: originals, sketches, stages, layers
[ ] Rights clause checked for "machine learning" / "technology development" licences

Questions and answers

Can I use AI in an art open call application?

It depends on the call. On 28 October 2026, 95% of open calls said nothing about AI, 61 banned AI-generated work, 8 required disclosure and 5 accepted it. If a call is silent, ask the organiser. Public funders like Canada Council and Arts Council England allow AI tools but hold you responsible for the content.

Do I have to disclose that I used ChatGPT for my artist statement?

Only if the call requires it. Using AI to translate or tidy your own text usually needs no disclosure, but every fact must be yours and checked. If the call asks for disclosure, say briefly what the tool did and what you wrote.

Can AI-generated art be copyrighted?

In the US, not without human authorship: Thaler v. Perlmutter (2025) confirmed this, and the Supreme Court declined the case in March 2026. The Copyright Office says AI-assisted work can be protected to the extent it shows human creative contribution; a prompt alone generally is not enough.

What if a jury wrongly thinks my work is AI-generated?

Show your process: original files, RAW photos, sketches, images of the work at different stages, layered files. Detectors are unreliable; process evidence is hard to fake. Keep it for every work you submit.