High Peak teens split 40/40 on whether AI art is cheating

A youth arts programme in the English Peak District ran a year-long project on AI and creativity, and the audience vote at its final debate is a more honest picture of public sentiment than most industry surveys. High Peak Community Arts published the results of “The AI Debate,” held in September 2025 at New Mills Festival. The numbers are striking: 40% of the audience said using AI to make creative work is cheating, 40% said it is fine as long as you are honest about it, and only 20% gave an outright no.

The project was run by the Level Up Programme, which supports young people aged 13 to 25 in the High Peak to build confidence and skills. Participants split into four groups, each working with a professional in a different art form: dance, creative writing, music, and visual arts. They created pieces exploring AI’s role in their discipline, joined by Josh Asquith, described in the source as a generative AI expert. The event combined live workshops, performance, and a panel discussion.

The results deserve attention from the AI industry, not because a small community event is statistically significant, but because it captures something the major labs keep missing. When OpenAI, Anthropic, and Google ship new image and music models, they frame the question as capability: what can the model do? The public is asking a different question: who gets to claim the result, and can any of it be trusted?

The ownership answer is the real signal

The clearest result in the debate was on ownership. 80% of the audience backed shared ownership between the user and the company running the AI model. 20% favoured Creative Commons for anyone. Nobody voted that the AI company or the user alone should own the work outright.

That is a remarkable consensus. It rejects both the maximalist positions that dominate the legal and policy fights. The AI companies, through their terms of service, generally claim broad rights to use inputs for training while granting users rights to outputs. The open-source community argues for no ownership at all, with everything released under permissive licenses. The courts are still working through whether training on copyrighted work is fair use, with cases like the Authors Guild suit against OpenAI and the Getty Images case against Stability AI still unresolved. The High Peak audience, mostly young people and their families, landed on a middle position that neither camp has articulated well.

Shared ownership is a pragmatic answer. It acknowledges that the model could not produce the work without the user’s prompt, direction, and curation. It also acknowledges that the company spent billions on compute and training data. Neither party can honestly claim sole authorship. The 80% figure suggests ordinary people intuit a joint-creator model that the current legal framework does not recognise.

Trust is the deeper problem

The second decisive result was on trust. 80% of the audience said they cannot trust AI on matters of opinion. 20% thought AI might offer a fresh perspective. Nobody voted yes.

This is the result the AI industry should worry about most. The capability conversation is about what models can do: write a sonnet, generate a plausible image, compose a melody. The trust conversation is about whether any of it means anything. A teenage audience in New Mills, after a year of hands-on experimentation, came out overwhelmingly sceptical that AI has anything useful to say about matters of judgement.

That scepticism is well founded. Large language models are trained to produce plausible text, not to hold genuine beliefs. They optimise for what sounds convincing, not for what is true. The industry has spent the last two years shipping models that are increasingly fluent and increasingly confident, and the public has noticed that confidence does not track accuracy. The High Peak result is a small data point in a larger pattern: the more people actually use these tools, the more they distrust them as authorities.

What the arts experiment shows about the tools

The four art-form groups produced different experiences of the same technology. The dance team drew inspiration from an abstract piece of art and explored questions of ownership through movement. The visual arts team collaborated with AI software to bring their drawings to life. The music team played live alongside AI-generated music, comparing the two. The creative writing team input characters to generate stories and used AI to read pieces back in their own voices.

The music comparison is the most instructive. Playing live music against AI-generated music in the same room is a direct test of what the technology can and cannot do. The audience heard the difference in real time. The creative writing team’s experiment of using AI to read their work back in their own voices is a genuinely interesting use case, one that points to a more constructive role for the technology: not as a creator but as a mirror.

These are the kinds of applications the industry does not talk about. The marketing focus is on autonomous agents and frontier models. The actual use, in a community arts programme in the Peak District, is more modest: a tool for exploration, a collaborator that can be directed, a way to hear your own words spoken back to you. The teenagers found the tools useful for process, not for product. That distinction matters.

What the industry should take from New Mills

The most encouraging number in the whole report is the last one. 60% of the audience left the debate saying something had shifted their thinking. That is the strongest argument for public engagement with AI that exists anywhere in the industry’s own communications.

The labs spend enormous sums on safety research, red-teaming, and policy papers. High Peak Community Arts spent a year running workshops with teenagers and produced a more honest conversation about AI’s social implications than most corporate AI summits. The format worked because it was hands-on, because it paired young people with working artists, and because it treated the technology as something to be touched and tested rather than feared or worshipped.

The parent feedback in the report is telling: “The event made me feel more positive about presenting and working through nerves.” That is not a comment about AI at all. It is a comment about what happens when people are given a structured, safe space to form and voice opinions about a scary technology. The confidence gains the programme reports, in critical thinking, public speaking, and forming arguments, are the actual deliverable.

The AI industry should study this model. The gap between what the public fears and what the public experiences is closing, but only where people get direct, guided exposure to the tools. The 40/40 split on cheating suggests the public is genuinely undecided, not hostile. The 80% on shared ownership suggests a workable compromise exists. The 80% on trust suggests the industry has a credibility problem that no benchmark release will fix.

The teenagers of High Peak have done what the industry has not: they have articulated a public position on AI authorship that is nuanced, practical, and shared. The question is whether anyone in San Francisco or London is listening.