As Artificial Intelligence systems consume creative assets, licensing is moving upstream from products to training data.
For years, the licensing business has centred on the end product, whether that is apparel, toys, publishing, entertainment, or promotional merchandise. But with the rise of generative AI, the conversation is moving upstream, and with it, the commercial foundations of intellectual property.
Licensing is no longer only about applying a character, brand, or creative asset to a finished product. Increasingly, it is also about whether that same intellectual property is being used to train the systems that generate new content, designs, and experiences.
Traditionally, a licensee would acquire rights to use specific IP on a product or within a defined category. Today, that same IP may also form part of the datasets and models that generative AI systems rely on to create material. That shift raises a very different set of questions around ownership, consent, compensation, and control.
This is where the issue becomes more complex. For AI developers, copyrighted works, characters, images, and brand assets may be treated as training data. For rights holders, however, they represent years of investment, creative development, and brand-building.
There is, of course, opportunity on both sides. Generative AI has the potential to open up new licensing models, new revenue streams, and new forms of collaboration between rights owners and technology platforms.
But the commercial tension is clear when material may already have been used to train systems without permission, payment, or even visibility, an issue that has been widely discussed in ongoing copyright and AI transparency debates.
For brand owners, that is the central concern. If your IP has value in the marketplace, it also has value in the training environment. The question is not only whether that value should be recognised, but how it can be recognised once a model has already ingested the material.
That is why the push for transparency, provenance tools, content labelling, and clearer disclosure around training data is gathering pace, particularly as the European Union advances transparency and copyright disclosure requirements through the EU AI Act framework.
At the same time, generative AI is making it increasingly difficult to distinguish between machine-generated output and original human-created work. For an industry built on ownership, authenticity and brand integrity, that is not a theoretical issue.
It goes directly to the value of licensed IP and the terms on which it is commercialised.
The speed of change is also highlighting how unsettled the market remains. In late 2025, OpenAI and The Walt Disney Company announced a three-year agreement to bring selected Disney, Pixar, Marvel and Star Wars characters to Sora for user-generated short-form video content, with some output expected for Disney+.
However, OpenAI later confirmed that Sora’s standalone web and app experiences would be discontinued, while subsequent industry reporting indicated Disney was reassessing its OpenAI-related AI video plans. Together, these developments show how fluid AI licensing strategies remain, even among the world’s largest entertainment companies.
There is still no settled playbook for AI licensing. Key questions remain around pricing, contractual structure, cross-border regulation, and the legal standards that may apply to the use of protected works in model training.
Those questions sit alongside broader enforcement challenges and growing regulatory complexity in major markets, particularly in the European Union and the United States where AI copyright and transparency rules continue to evolve.
What is already clear, however, is that waiting for certainty is not a strategy. The companies likely to be best positioned are those already reviewing their IP portfolios, updating agreements to address AI-related rights, and exploring direct partnerships with developers and platforms.
AI is not replacing the licensing industry. But it is changing the terms on which that industry operates. For rights holders, agents, and licensees alike, the focus is shifting from simply protecting IP to actively managing, monetising, and enforcing it within AI ecosystems.
That shift is no longer theoretical. It is already underway. Companies are reassessing how IP is protected, valued, and commercialised in an AI-driven environment, with AI rights, training data, attribution, and control becoming part of modern licensing strategy.
This article also appeared in Edition 54 of The Bugg Report Magazine
