Telescope Logo
Telescope Logo

The content licensing reset

Read time: 5 minutes


For three years, the AI and copyright debate has been told as a battle: developers against creators, innovation against protection. That framing is becoming outdated.

The law is far from settled, but its direction of travel is increasingly visible: less a question of whether AI may use content, more a question of on what terms and at what price. For platforms, AI developers and rightsholders alike, the strategic question is shifting from “is training lawful?” to “what is the right deal?”

Man working on a computer

For three years, the AI and copyright debate has been told as a battle: developers against creators, innovation against protection. That framing is becoming outdated.

The law is far from settled, but its direction of travel is increasingly visible: less a question of whether AI may use content, more a question of on what terms and at what price. For platforms, AI developers and rightsholders alike, the strategic question is shifting from “is training lawful?” to “what is the right deal?”

Market shift: from the right to prohibit to the right to be paid

Copyright was built on an exclusive right to authorize or prohibit, exercised work by work. Generative AI is putting pressure on that model. Training corpora are measured in billions of items, and individual negotiation, or individual proof of use, struggles to operate at that scale. What is emerging instead is a logic of volume: remuneration rather than prohibition, and collective or standardized mechanisms rather than bespoke permissions. The market is already moving in that direction. Licensing deals between AI developers and publishers have multiplied, and their structure is evolving from one-off lump sums toward usage-based models linked to grounding or to the contribution of content to AI-generated answers. Large litigation settlements are also beginning to act as informal price references. None of this resolves the underlying legal questions, but it shows that the parties themselves are treating content as an input to be priced rather than a battleground to be won.

Abstract blue backrgound

Licensing deals between AI developers and publishers have multiplied, and their structure is evolving from one-off lump sums toward usage-based models

Market shift: from the right to prohibit to the right to be paid

Copyright was built on an exclusive right to authorize or prohibit, exercised work by work. Generative AI is putting pressure on that model. Training corpora are measured in billions of items, and individual negotiation, or individual proof of use, struggles to operate at that scale. What is emerging instead is a logic of volume: remuneration rather than prohibition, and collective or standardized mechanisms rather than bespoke permissions.

The market is already moving in that direction. Licensing deals between AI developers and publishers have multiplied, and their structure is evolving from one-off lump sums toward usage-based models linked to grounding or to the contribution of content to AI-generated answers. Large litigation settlements are also beginning to act as informal price references. None of this resolves the underlying legal questions, but it shows that the parties themselves are treating content as an input to be priced rather than a battleground to be won.

Abstract blue backrgound

Licensing deals between AI developers and publishers have multiplied, and their structure is evolving from one-off lump sums toward usage-based models

Global legal landscape: Europe searches for a framework

The EU: harmonization first. Europe is where the debate is most structured, and where its outcome is least certain. The 2019 Digital Single Market Directive allows text and data mining (TDM) subject to rightsholders’ ability to opt out. Whether, and how far, that exception covers the training of large language models is now the central open question. On March 10, 2026, the European Parliament adopted a resolution encouraging voluntary, sector-wide collective licensing and inviting the Commission to review whether the copyright framework is fit for generative AI. National initiatives illustrate the limits of going alone. In France, the Senate adopted the “Darcos” bill in April 2026, creating a presumption that protected works had been used in training. The text has since stalled: it raised serious questions of practical feasibility, since proving that a given work was not used may be technically impossible, and of compatibility with an EU regime in which Member States cannot add to the list of copyright exceptions. The French parliamentary work that followed points instead to EU-level solutions, such as a European register of opt-outs and a trusted third party. The lesson is broader than France: in a cross-border digital market, isolated national rules risk fragmentation, and the real decisions will be taken in Brussels and Luxembourg.

The CJEU: the decisive forum. Two references from Hungarian courts will shape the EU-wide answer. Like Company v Google (C-250/25) asks whether training a chatbot and generating its outputs involve acts of reproduction or communication to the public, and how the TDM exception applies. The more recent Manchester United Fanatics Klub v Google (C-806/26) raises questions that go to the heart of how AI services operate in practice: the status of synthesized answers that draw on identifiable sources, the processing of user prompts, grounding and whether the absence of an express opt-out may amount to consent. The Court’s answers could redefine the TDM exception, and the role of opt-out mechanisms, across the Union.

On both sides of the Atlantic, the issues converge on the same points: how content was acquired; what the model reproduces, retains, and has access to; and whether a market for licensing exists.

Germany and the United States: different routes, similar questions. German courts have been among the first in Europe to address copyright and generative AI. For example, the Munich Regional Court held at first instance that protected lyrics had been reproduced through recoverable model retention and in generated outputs. The judgment is under appeal and should not be read as a general finding that AI training is unlawful. It nevertheless highlights three questions that any license should address: what may be used for training, what the model may retain and what it may reproduce in its outputs. In the United States, early rulings have leaned toward findings of fair use for training on the facts before them, while leaving open the treatment of unlawfully acquired data and of market harm. As one judge put it, if using copyrighted works to train models is as necessary as the companies say, “they will figure out a way to compensate copyright holders for it.” On both sides of the Atlantic, the issues converge on the same points: how content was acquired; what the model reproduces, retains, and has access to; and whether a market for licensing exists.

Operational impact: content becomes a line item

Whatever the final legal answers, legal uncertainty itself now has a cost. For AI developers and platforms, data provenance is becoming a governance and balance-sheet issue as much as an engineering one: documenting sources, respecting machine-readable reservations and being able to explain how models are trained and deployed. Transparency obligations under the AI Act add to that expectation. For rightsholders, the ability to show clear rights chains and validly expressed opt-outs is becoming a commercial asset. For platforms, which often both supply content and deploy AI, the picture is two-sided: they are licensors and licensees at once. Their interest lies in a market that is predictable, scalable and harmonized across Europe, rather than one shaped by divergent national rules and case-by-case litigation.

Code on a computer screen

Operational impact: content becomes a line item

Whatever the final legal answers, legal uncertainty itself now has a cost. For AI developers and platforms, data provenance is becoming a governance and balance-sheet issue as much as an engineering one: documenting sources, respecting machine-readable reservations and being able to explain how models are trained and deployed. Transparency obligations under the AI Act add to that expectation. For rightsholders, the ability to show clear rights chains and validly expressed opt-outs is becoming a commercial asset. For platforms, which often both supply content and deploy AI, the picture is two-sided: they are licensors and licensees at once. Their interest lies in a market that is predictable, scalable and harmonized across Europe, rather than one shaped by divergent national rules and case-by-case litigation.

Code on a computer screen

Opportunities: what to negotiate now

Parties that negotiate now can help shape emerging benchmarks rather than inherit them. Platforms and AI developers will want to consider:

  • Clear scope: Licenses that address the different stages of use, from building datasets and training to grounding and outputs.
  • Legal certainty for the past: Releases or covenants covering historic use.
  • Balanced risk allocation: Warranties on rights chains and opt-out status, and consistent back-to-back terms where third-party models are used.
  • Flexibility: Review or repricing clauses tied to the forthcoming CJEU rulings and any EU reform.
  • Scale: Collective or pooled arrangements that reduce transaction costs.

Parties that negotiate now can help shape emerging benchmarks rather than inherit them. Platforms and AI developers will want to consider:

Key takeaways

  1. The law is moving towards remuneration rather than prohibition. Legislators and courts are increasingly asking how content should be paid for, not whether AI may exist. But the rules are still being written, and caution is warranted.
  2. Europe will decide at EU level. National initiatives, such as the French bill, show the limits of unilateral action; the CJEU and the Commission’s review are where the framework will be set.
  3. Deal structure matters more than headline fees. Scope, treatment of past use, risk allocation and review clauses will determine long-term value. Businesses that build flexible, well-documented licensing strategies now will be best placed whatever the final rules turn out to be.

Key takeaways

  1. The law is moving towards remuneration rather than prohibition. Legislators and courts are increasingly asking how content should be paid for, not whether AI may exist. But the rules are still being written, and caution is warranted.
  2. Europe will decide at EU level. National initiatives, such as the French bill, show the limits of unilateral action; the CJEU and the Commission’s review are where the framework will be set.
  3. Deal structure matters more than headline fees. Scope, treatment of past use, risk allocation and review clauses will determine long-term value. Businesses that build flexible, well-documented licensing strategies now will be best placed whatever the final rules turn out to be.

Authors


Umar Bakhsh

Umar Bakhsh

Partner, US

View profile
Caroline Lyannaz

Caroline Lyannaz

Partner, Digital Content Sub-Sector Co-Lead, France

View profile
Nils Muller

Nils Müller

Partner, Digital Content Sub-Sector Co-Lead, Germany

View profile
Louise Mouclier

Louise Mouclier

Senior Associate, France

View profile

In this edition


In this edition


Children on phones

Age assurance as digital infrastructure

Read time: 7 minutes

Read more

Illegal content in the AI supply chain

Read time: 6 minutes

Read more

Platform governance across multiple rulebooks

Read time: 5 minutes

Read more

AI agents and consumer liability: new interfaces, new risks

Read time: 8 minutes

Read more

Creators, advertising, and monetization

Read time: 6 minutes

Read more

Protecting premium content at internet speed

Read time: 6 minutes

Read more

Disputes and enforcement in the digital content sector

Read time: 4 minutes

Read more

Previous page

Next page

Arrow

© Eversheds Sutherland. All rights reserved. Eversheds Sutherland is a global provider of legal and other services operating through various separate and distinct legal entities. Eversheds Sutherland is the name and brand under which the members of Eversheds Sutherland Limited (Eversheds Sutherland (International) LLP and Eversheds Sutherland (US) LLP) and their respective controlled, managed and affiliated firms and the members of Eversheds Sutherland (Europe) Limited (each an "Eversheds Sutherland Entity" and together the "Eversheds Sutherland Entities") provide legal or other services to clients around the world. Eversheds Sutherland Entities are constituted and regulated in accordance with relevant local regulatory and legal requirements and operate in accordance with their locally registered names. The use of the name Eversheds Sutherland, is for description purposes only and does not imply that the Eversheds Sutherland Entities are in a partnership or are part of a global LLP. The responsibility for the provision of services to the client is defined in the terms of engagement between the instructed firm and the client.

Connect with us

linkedin logo
facebook icon
youtube icon