As governments around the world consult on the use of copyrighted works to train Large Language Models (LLMs), Cambridge University Press & Assessment has been pushing for licensing arrangements that would protect creative industries worldwide. In the last 12 months, Cambridge has responded to consultations with the Indian government, the EU and the UK government.
CAMBRIDGE – Cambridge University and other leading voices, including the Publishers Association in the UK, have been calling for equitable use of copyrighted data by AI models. This includes transparency around the copyrighted content that has been used to train LLMs, fair remuneration for this use and clear attribution of source material to ensure that copyright owners receive recognition for their work (and which helps develops trust in the validity of AI generated content), according to the official website of the University.
In February 2025, Cambridge responded to a UK government proposal that could give AI companies unrestricted access to copyrighted work unless owners proactively ‘opt out’, warning that the risks of this approach would allow tech companies to scrape copyrighted material, including academic research and learning materials, without sufficient protection. The response highlighted that effective, proven licensing solutions are already in place and being used by some AI companies.
In the last few weeks, consultation responses have been submitted to the European Union (EU) and to the Government of India’s Working Paper on Generative AI and Copyright.
The EU consultation focused on how the 2024 EU Artificial Intelligence Act, widely regarded as the most comprehensive regulation of AI, is being implemented in practice. It specifically focused on the current systems available for copyright owners to opt out from their works being used to train LLMs.
Cambridge’s response raised doubts about the effectiveness of these systems, which are not currently robust enough to ensure that copyrighted works are not ingested for AI training purposes.
In India, the government is consulting on a proposal to strike a balance between AI innovation and copyright. A central part of this is a ‘Hybrid Model’, a blanket license giving access to all copyrighted content in exchange for a single flat fee, with copyright owners receiving a share of this revenue.
The stated aims of the Hybrid Model are to simplify licensing procedures, reduce transaction costs and ensure fair access to content for training AI systems.
In its response to the consultation, however, Cambridge argued that not only would these aims not be realized but that the Hybrid Model would effectively upend the foundations upon which copyright law is based, causing harm to the creative industries in India and elsewhere.
The response instead advocated for the adoption of existing collective and individual licensing solutions that have already proven effective. This would best support India’s ‘Orange Economy’, a model adopted in 2025 that is focused on supporting the country’s vibrant creative industries.
Arun Rajamani, Managing Director at Cambridge University Press and Assessment, South Asia, said:
“Ethical and responsible AI development depends on ensuring that creators and rights holders are appropriately recognized and compensated when their works are used to train Large Language Models.
“As policy discussions continue, it is important to consider whether proposed approaches achieve this objective in a balanced and effective manner.
“Above all, efforts to support AI development should seek to preserve the rights of creators and rights holders, whose works are integral to the development of high-quality, reliable AI technologies.
“As an education provider and publisher with deep connections across India, Cambridge recognizes the potential for responsibly developed AI systems to deliver meaningful benefits to society. Ensuring fair recognition and compensation for creators remains an important component of that process.”
Mandy Hill, Managing Director at Cambridge University Press, and President of the Publishers Association, said:
“Many countries around the world benefit enormously, both culturally and economically, from the output of their creative industries.
“But these industries are threatened by the refusal of big tech companies to adopt existing, effective licensing arrangements for content that has or is being used to train Large Language Models.
“Cambridge supports innovations in AI, responsible licensing, and collaboration with the tech industry. Key to achieving this balance is for governments to uphold the legal foundations of copyright when developing AI policies to avoid damage to the creative industries.”
Catie Sheret, General Counsel at Cambridge University Press & Assessment, said:
“While legislative and regulatory debates continue, it is important to recognize that copyright already provides meaningful safeguards for creators and content owners: in most key markets, large‑scale commercial text and data mining of protected works is not a ‘free‑for‑all’ but is constrained by existing rights, opt‑out mechanisms, contract terms, and database protections. The real problem for today is not a lack of rights, but the practical and legal friction these differing rights create. We are advocating for a considered, fair, coherent approach based on the existing copyright frameworks in place around the world.”