I still remember the early days of telecom liberalisation, sitting in meetings with operators who refused categorically to share their infrastructures.
They owned the best rooftops, the best towers, the best fibre routes and the best ducts and central locations.
And they sincerely believethat sharing these assets would destroy their competitive advantage. Some, and not only incumbents, were almost offended at the idea. They saw mutualisation as a threat, a loss of power, an erosion of identity.
Fast forward twenty years: what once felt unthinkable has become natural.
Open access is routine. Passive co-location is expected. Fibre mutualisation is efficient. Wholesale access is regulated.Interconnection is symmetrical. Infrastructure sharing is simply part of how modern networks function.
The very operators who once fought against opening their networks now consider it an obvious and rational practice. That is because in network industries, value does not come from exclusivity but from externalities. The more people participate, the more everyone benefits.
And today, as AI becomes a fundamental cognitive infrastructure, we face the exact same logic with copyright.
1. Luxury goods vs. network goods: two different economics
In consumer markets, price often follows function. But luxury goods are an exception. Their price reflects symbolism, aspiration. It is in a way the price of the dream.
Network goods follow the opposite principle. Their value grows with participation.
Telecoms, payments, transport networks, cloud infrastructures: their utility depends on scale, interconnection, and collective adoption.
AI belongs unquestionably to the network category.
2. Copyright is still built on a scarcity logic
Traditional copyright assumes information is scarce. You needed to buy the newspaper. You waited for the 8 p.m. evening news. You depended on gatekeepers to access knowledge.
Copyright protected this scarcity. It created rents built on control.
Today, that world no longer exists. Information is abundant, real-time, user-generated, accessible 24/7, cheap or free.
The scarcity that justified the copyright logic has evaporated.
We have moved from a world of rationing information to a world of permanent informational overflow.
3. The lesson from telecoms: opening the network increases total welfare
My career in network regulation taught me a clear lesson: when externalities are positive, exclusivity becomes inefficient. Opening access increases the collective value of the system.
This is why regulators introduced local-loop unbundling, passive mutualisation, duct sharing, fibre co-investment models, symmetric interconnection, non-discrimination obligations, and functional separation.
These reforms were resisted, then adopted…then embraced. Because a network is not a luxury good. Its power lies in participation.
AI is no different.
4. AI behaves like a cognitive infrastructure, not a private artefact
AI’s performance is shaped by billions of texts, millions of images, code repositories, public conversations and human culture at scale. It improves with diversity, scales with participationI and grows through feedback loops.
AI is structurally a collaborative network, not a discrete product.
Trying to govern AI with a copyright logic rooted in scarcity is as incompatible as trying to regulate telecommunications by restricting interconnection.
5. The economy is shifting: from competition to collaboration
The modern digital ecosystem is already built on collaboration:
• Open APIs
• Interoperability
• Open-source foundation
• Shared cloud infrastructure
• Multi-stakeholder governanc
• Transformative fair use in the U.S….
Our economic reality has changed. Copyright must change with it.
6. Copyright must evolve from exclusivity to governance
In an AI environment, copyright should no longer act as a barrier to access. It must become a governance interface, enabling attribution, transparency, responsible access, interoperability, value sharing, and social contribution.
The objective of our paper is not to weaken copyright but to adapt it to the reality of a network-based intelligence system.
7. A voluntary and ethical solution: a pure Opt-In system
One symbol. One decision. Full respect for creators.
Rather than forcing creators into AI training or obliging them to “opt out” through complex legal processes, the future should rest on a single, clear, respectful mechanism:
A pure Opt-In mark for AI training. How it works
- The creator explicitly decides that their work may be used to train AI.
- They display a public symbol : a simple, verifiable mark.
- Anyone without the symbol remains fully protected under traditional copyright.
This is the cleanest, most transparent solution. It respects freedom, creators and the nature of AI as a network good.
8. The “AI-Responsible” mark
A voluntary label like « Eco-Friendly », « Fair trade », or « Creative commons ».
The Opt-In mark becomes a signal of contribution. A badge of responsible participation in the collective intelligence ecosystem.
The message is simple: “I allow my work to be used for responsible AI training.”
This label creates clarity for users, certainty for developers, empowerment for creators, and legitimacy for the entire governance model.
Copyright doesn’t disappear, it is reinterpreted as a right to contribute.
… The only barrier to entry that must remain high: personal data
As information becomes infinite and AI becomes generative, the only legitimate scarcity left is privacy.
In Europe, GDPR, the Charter of fundamental rights, and the AI Act protect personal data as a matter of dignity and autonomy.
In the U.S, CCPA, CPRA, and state privacy laws follow the same principle.
The last real barrier to entry is not creative works but the protection of individuals and their freedom rights.
About the author Zied Miled is an engineer-lawyer specializing in network industries, regulation, and AI. The founder of Aidalet and Doudy.ai, He designs native-AI legal and educational technologies that enhance human judgment. He advises institutions across Europe, the MENA region, and Türkiye on regulatory transformation and AI ethics.ıı