Morgane Oger

The Public Has Rights in Culture Too

CanadaLaw and SocietyTechnology
The Public Has Rights in Culture Too

This started with Kris Krüg’s post, “Artists Learn. Machines Extract.”, responding to my question in a LinkedIn thread about whether machine training on artists’ work is categorically different from artists learning from artists.

Kris argues that the answer lies in “consent, scale, secrecy, and substitution.” He frames commercial AI training as converting culture into private infrastructure – ultimately, a “mine.” I think this analogy fits to a meaningful degree, especially for large-scale commercial extraction by foreign actors operating without adequate oversight or benefit-sharing. It highlights real risks of enclosure and power concentration. At the same time, it leaves something important out.

Centring the Producer – and the Public

The AI-and-art debate often centres the creator whose work is used. That focus is necessary: creators have legitimate rights and interests. Yet it frequently overlooks the user of the work, the learner, the builder, the public, and the commons that made the original work valuable in the first place.

Copyright protects expression – the specific, original form in which ideas are fixed. It does not protect ideas, facts, techniques, themes, styles, influence, or the act of learning from a work. As Canada’s Intellectual Property Office (CIPO) explains, ideas, facts, short titles, unfixed works, and works lacking original skill and judgment fall outside copyright protection. Copyright safeguards original expression fixed in a material form.

This distinction is central to the AI debate.

Technological Change Is Not New

Artists have already navigated waves of technological displacement – from physical media to digital tools like layers, filters, brush engines, vector paths, and automated workflows. The market absorbed these shifts through adaptation and professional practice.

Much of that creative tooling rests on layers of software labour – some commercial, much open source, and a great deal contributed by standards bodies, volunteer maintainers, academics, community documentation, unpaid bug reports, shared scripts, and invisible technical conventions.

The Technical Commons We Already Rely On

Open-source licences explicitly permit redistribution, modification, derived works, and commercial use. Software developers contribute to the commons for practical, reputational, and ideological reasons. Creative professionals do something similar when they publish online, accepting that their work will be seen, interpreted, classified, compared, and built upon.

Many artists and creative workers have long depended on this technical commons. Now that parts of creative production are becoming automatable, some who benefited from it understandably feel threatened. Change is difficult – software developers have experienced this for decades.

The Asymmetry in the Conversation

There is a notable asymmetry: creative industries have routinely benefited from unlicensed software, cracked tools, open-source libraries, and unpaid technical infrastructure, while software labour’s contributions are rarely credited. If we are serious about consent, extraction, and cultural labour, the conversation must include the builders of the digital foundations that make modern creative practice possible.

AI Scrapes Everyone – and Raises Sovereignty Questions

AI systems scrape my work too – software, writing, advocacy, code, public records, images, and the ordinary residue of online life. This is now a shared condition.

The deeper issue is that much current frontier AI development is driven by foreign commercial actors for private profit, using Canadian (and global) data, culture, and labour without sufficient local governance or accountability. Kris’s “mine” framing resonates here: when extraction serves private enclosure at industrial scale, it risks turning public culture into privatised infrastructure.

This pattern is global. Every community faces the question of who gets to absorb, model, profit from, and sell back its knowledge.

Sovereignty and Public Benefit

Canada is responding. The Canadian Sovereign AI Compute Strategy invests in domestic capacity to safeguard data and IP while supporting made-in-Canada solutions. The June 2026 AI for All strategy emphasises trust, opportunity, and sovereignty.

If AI is trained on all of us, the core public-interest question is whom the trained system ultimately serves.

A privately owned model returning private rent raises serious concerns. A sovereign, accountable, public-serving AI – advancing education, accessibility, public safety, language preservation, climate adaptation, health, justice, and democratic resilience – offers a different path.

Sovereign states could explore mechanisms such as charging for access to our commons and sovereign intellectual wealth. This could provide accountability and benefit-sharing. At the same time, such tools cut both ways: others are likely to apply the same approach to us, so any framework must be carefully designed to avoid escalation or fragmentation of the global knowledge commons.

This does not erase copyright, privacy, consent, attribution, or accountability. Privacy commissioners rightly demand legal authority, appropriate purposes, and transparency.

Human Learning, Machine Learning, and Cultural Inheritance

Humans are themselves compositions of physical processes leading to consciousness, intent, and creative work. AI systems are not persons, but as they imitate reasoning and creativity, we must scrutinise our assumptions.

If education is a public good because society benefits when minds are trained on civilisation’s record, then public-serving AI should not be categorically denied the same cultural inheritance – especially when governed locally and accountable to the communities it serves.

This does not mean “anything goes.” We must distinguish extraction for private enclosure from learning for public benefit; copying protected expression from learning patterns; market substitution from building public capability; and foreign dependency from sovereign infrastructure.

Precision Over Polemics

Kris is right that frontier AI development deserves scrutiny on scraping, retention, substitution, transparency, displacement, and power concentration. Canada’s copyright consultations reflect the ongoing split between creator concerns (consent, credit, compensation) and arguments that text and data mining often targets facts and patterns rather than protected expression.

The legal and ethical line should focus on whether protected expression is reproduced, retained, exposed, substituted, or passed off. Artists have always learned from predecessors through influence, reuse, and transformation. AI changes the scale, speed, access, and economics – serious shifts, but not automatically theft.

AI is aggregate pattern recognition from knowledge placed into circulation. It is trained on all of us. This makes it a sovereignty issue as much as a copyright one.

The artist remains protected in specific expression – the poem, drawing, photograph, or final work carrying human judgment. AI can automate mechanical tasks, but art still requires spark, taste, intention, context, risk, and human meaning-making.

The Better Question

When claims arise that “AI stole from artists,” we need precision: Did it copy protected expression? Reproduce a substantial part? Output something confusingly similar? Violate licences or bypass controls? Substitute for a specific market?

A broader prohibition on machines learning from public culture would grant creators power copyright was never meant to confer – control over interpretation, pattern recognition, automation, and learning itself.

The public has rights in culture too.

The distinction is not (or should not be) purely anthropomorphic. Human learning inside a head is treated as noble; machine learning is often cast as extraction. Artists using open tools and unpaid infrastructure are creators; AI using public patterns is sometimes labelled thief.

The better question is: What was copied? What was protected? What was substituted? What was concealed? Who benefited? Who was harmed? Who now controls the intelligence built from all our work? And does the law preserve a fair balance between creators, users, builders, communities, and the public?

Receipts / References