In 2025, an amateur musician can compose an album in one afternoon. Millions of others do the same. MONDIACULT 2025 was held in Barcelona from September 29 to October 1, 2025 and brought together participants from 163 countries. It called for transparent AI and the protection of creators’ rights and contributions, without explicitly adopting a consent and remuneration mechanism for model training.
The Essential Points
- Generative AI is exploding cultural supply without demand following suit, compressing creators’ revenues across all sectors.
- MONDIACULT 2025 approved recommendations on consent, remuneration, and algorithmic transparency, but no state has translated them into binding obligations (UNESCO, independent expert group report, 2025).
- Recommendation algorithms concentrate visibility on global hits: diversity of supply does not translate into diversity of consumption, according to the European Audiovisual Observatory.
- National regulation hits a structural wall: platforms operate outside the jurisdictions where their effects are felt.
- The stakes extend beyond culture: if Barcelona’s recommendations remain dead letters, it is an entire global digital governance that draws a lesson from this.
The Abundance That Impoverishes
No general international mechanism imposes creator remuneration for training, while the European Union has established binding obligations for transparency and copyright compliance. Cultural consumption remains concentrated on a restricted number of global titles, despite access to millions of works, according to the European Audiovisual Observatory. AI lowers production costs. Certain recommendation algorithms reinforce visibility inequalities through a popularity bias, depending on the platforms and systems studied.
An increase in supply can intensify competition for attention and affect the value of certain works, without mechanically causing a drop in the value of each individual unit. Generative AI imported this into culture at a speed few had anticipated. Spotify reports more than 100,000 new songs published daily; the claim of a doubling in three years is not established. Music generated by AI represents a growing share of these uploads, often without identifiable human attribution. The arithmetic result is brutal: with stable total listening hours, each title captures a smaller share of the pie.
The logic extends beyond music. In publishing, titles can be generated automatically. In imagery, stock banks like Getty or Shutterstock had to resolve a new question: whether or not to accept AI content, with effects on the rates offered to human photographers. YouTube receives more than 400 hours of video per minute, or approximately 17 days of continuous viewing for content uploaded in just one minute alone.
These figures do not herald the end of human creation. They indicate that its economic conditions are changing radically and rapidly. An increase in supply can intensify competition for attention and affect creators’ revenues.
The Algorithm as Cultural Prescriber
The European Audiovisual Observatory notes visibility challenges and a concentration of promotional slots on platforms. Spotify, YouTube, or Netflix’s algorithmic recommendations can favor already popular content. These algorithms optimize short-term engagement—that is, listening or viewing duration. Already popular works generate more reliable engagement data than emerging creations, which feeds their visibility and marginalizes others.
The phenomenon can be self-sustaining. An artist’s visibility in recommendation systems varies depending on platforms and their criteria. The digital cultural market can concentrate attention on a limited number of contents.
Algorithmic governance and investigative journalism share the same mechanism: platforms capture audiences and redistribute value according to their own rules, without obligation to account for their editorial choices. For culture, the volume at stake is incomparable to that of journalism.
The Barcelona Decision and Its Immediate Effects
MONDIACULT 2025 is the largest UNESCO global conference on cultural policies since MONDIACULT 2022, held in Mexico. MONDIACULT 2025 brought together participants from 163 countries; the independent CULTAI group report fed debates, while ministers adopted a separate outcomes document. The report advocates for more equitable cultural economies and prevention of unfair use of works, without a determined value-sharing mechanism being confirmed in the official summary.
These recommendations are politically significant. They represent a rare multilateral consensus on a subject where positions between states, industries, and generations of creators diverge strongly. They also constitute a reference framework that national or regional legislators can grasp.
Less than a year after MONDIACULT 2025, no general international remuneration mechanism yet exists; however, the EU already applies binding transparency and copyright obligations to GPAI model providers. The European Union has the AI Act and the Digital Markets Act, two texts whose effects on creator remuneration remain limited and partial. The United States has no comparable federal legislation. And platforms, whose servers and headquarters are distributed across different jurisdictions, can sometimes absorb isolated national injunctions without substantially modifying their model.
Cultural Rent as a Laboratory for Platform Capitalism
Bruno Patino, editorial director of Arte and author of The Time of Human Obsolescence (2026), articulates his thesis around three notions: attention economy, inundation by the flow, and programmed obsolescence of human creators. The algorithmic markets for music or imagery provide the most quantifiable illustration. Rights holders contest unlicensed uses of their works for training and demand remuneration; the overall effect on creators’ revenues is not established by this observation alone. The distribution of visibility by algorithms rests on criteria specific to platforms.
Patino associates generative AI with transformations in platform capitalism. MONDIACULT 2025 calls for better protection of creators’ rights and contributions in the AI context; it does not note that all models have been trained without consultation or remuneration.
The competing reading deserves serious consideration. Liberal economists, such as Tyler Cowen, have argued that AI acts as an equalizer by enabling resourceless creators to achieve production quality previously reserved for well-funded studios. The argument holds on the production front: technical costs indeed collapse. But it does not resolve the question of value distribution: producing better is insufficient if audience access remains controlled by algorithms that structurally favor already established content. AI as a tool for creator empowerment does not automatically imply better remuneration if the market where they sell remains asymmetrical.
The equalization argument perhaps applies better to emerging cultural markets, where creators from low-income countries can now produce with means comparable to those of major studios. This is a reality. It does not erase the other: access to production does not guarantee access to audience.
Barcelona in 2030
The MONDIACULT 2025 moment is a test for global digital governance, not only for culture. If recommendations from a conference organized by UNESCO and bringing together 163 countries remain without concrete translation in the five years following, this practically invalidates a model of multilateral regulation for the digital sphere. The signal would be received far beyond culture.
Several trajectories are plausible by 2030-2035. The first rests on bloc-based regulation. The European Union, drawing on its precedents with the GDPR and AI Act, could impose within its internal market obligations for consent and value-sharing that apply to platforms wishing to operate in Europe, regardless of server location. This extraterritorial approach has partially worked for personal data; it could work for cultural rights, provided member states coordinate and sanctions are sufficiently dissuasive. The GDPR precedent suggests that platforms adapt when the market at stake is large enough to justify compliance.
The second trajectory is that of the market itself. Platforms like Bandcamp, SoundCloud, or more recent entrants have built models that remunerate creators differently, sometimes better. If a portion of listeners and consumers deliberately chooses these platforms for ethical, quality, or diversity reasons, the market can generate pressure without awaiting the regulator. This trajectory is slow and uncertain, but signals exist: the rise of direct creator subscriptions via Patreon or Substack suggests a fraction of the public is prepared to pay directly, bypassing algorithms.
The third trajectory, less favorable, is that of fragmentation. Each major regional bloc adopts its own rules—the European Union for consent, China for cultural sovereignty, the United States for free market—and platforms adapt their behavior on the surface without modifying their deep architecture. In this scenario, Barcelona’s recommendations become a symbolic reference, cited in speeches but without operational force. This is the risk most documented by precedents in digital governance from the 2010s.
What will distinguish these trajectories in coming years rests on a few observable signals. The European Union’s capacity to impose algorithmic transparency obligations on platforms under the Digital Services Act, decisions by American and European courts on ongoing lawsuits around copyright and AI model training, and whether or not interoperable technical standards emerge allowing creators to control use of their works—these three elements will sketch, by 2027-2028, which scenario prevails.
Levers Available to Creators Before Any State Decision
Awaiting regulation is not the only option available. Several mechanisms are developing that allow creators to partially regain control, independent of legal framework. Technical licenses that mark works to indicate conditions for use by AI models, similar to what Creative Commons did for traditional copyright, are advancing in several consortiums. The Content Authenticity Initiative, led by Adobe and media partners, develops attribution standards that allow tracing a work’s origin even after modification. These tools do not replace legal obligation, but they create an infrastructure that can facilitate application when regulators act.
Collectives of creators in Europe are experimenting with cooperative remuneration models, where revenues generated by algorithms are pooled among members rather than captured individually. In Germany, GEMA, the society managing musical rights, has engaged in direct discussions with AI developers on license agreements covering model training. In France, SACEM has followed a comparable approach. These bilateral agreements are imperfect and cover only part of uses, but they constitute contractual precedents that could feed future regulation.
The question of worker autonomy in an algorithmic environment applies to creators as it does to employees: tools change faster than rights. The difference is that creators have, historically, greater experience with collective organization to defend these rights, and the legal precedents they are building today around AI will be used far beyond the cultural sector.
AI is already transforming culture. Barcelona leaves another question open: are democratic societies capable of imposing rules on global private infrastructures whose effects are felt everywhere, but whose governance remains concentrated in few hands. Culture constitutes the most visible experimental ground for this challenge. The answers produced in the next five years will inform the capacity of democracies to regulate digital capitalism.
Sources
- UNESCO, Report of the Independent Expert Group on AI and Culture, MONDIACULT 2025, Barcelona: https://www.unesco.org/en/articles/new-expert-report-explores-how-ai-transforming-culture
- Bruno Patino, The Time of Human Obsolescence, 2026: https://www.lalibre.be/culture/livres-bd/2026/05/21/nous-approchons-de-la-fin-possible-de-lhumain-RMTSEKPNTJFHDDY5MHZY577CSE/
- European Audiovisual Observatory, reports on European audiovisual markets and cultural diversity (Council of Europe, Strasbourg)
- Content Authenticity Initiative, attribution standards for digital content (Adobe and partners)
- GEMA (Gesellschaft für musikalische Aufführungs- und mechanische Vervielfältigungsrechte), discussions on AI licenses for model training
