CABI Blog

This blog is written by Damian Bird, Publishing & Operations Director at CABI, and explores how publishers can engage with generative AI in ways that are ethical, transparent and economically sustainable, with a particular focus on valuing and licensing trusted knowledge.

Generative AI is moving fast, and publishers are feeling the effects. Content that once circulated through established research, educational and practitioner channels is now being incorporated into AI systems. These systems can summarize, reframe or reproduce knowledge at scale.

This shift brings real risks: unlicensed reuse of copyrighted material, distortion of peer‑reviewed guidance, market substitution, and blurred accountability when things go wrong.

But it also brings opportunity. Responsibly designed AI tools can extend the reach of trusted knowledge, support better decision‑making, and lower barriers to access, particularly in agriculture, environment and development. The challenge for publishers and their authors is therefore how to engage with AI in ways that are ethical and economically sustainable.

As part of the ‘Generative AI for Agriculture (GAIA)’ project, funded by Gates Foundation and UK International Development from the UK government, CABI has been working with partners to explore what ethical, transparent and sustainable engagement with AI looks like in practice.

Central to this work is the development of pricing guidance for AI content licensing, developed in collaboration with the AI licensing consultancy, IndieLex.

At its core is a simple principle: sustainable AI depends on sustainable ways of valuing trusted knowledge.

Pricing has become strategic

Much of today’s AI ecosystem was built on large‑scale web scraping, often justified by broad interpretations of “fair use” or claims that AI outputs are sufficiently “transformative”. Those assumptions are now being challenged by courts, by publishers, and by AI developers themselves.

In this context, pricing is about more than income. It enables publishers to assert that expert‑authored and reviewed content is not a free or interchangeable commodity. It also helps define acceptable uses of their material and create accountability, traceability and governance around AI reuse. In doing so, it supports a shift from reactive enforcement to proactive, licensed engagement.

Pricing frameworks help articulate the value of curated, human‑authored knowledge to AI systems. This is particularly important as the online environment becomes saturated with low‑quality or synthetic data.

One price does not fit all

A key insight from the GAIA project is that AI licensing is not a single market. Different uses of the same content carry very different levels of value and strategic implication.

There is a clear difference between content used to train AI models in a way that permanently shapes how they perform; content used for retrieval‑augmented generation (RAG) or fine‑tuning, where access is more controlled and time‑limited; and content used internally for research compared with content embedded in commercial, user‑facing AI products.

Recognizing these differences allows publishers to make more nuanced decisions. It supports stronger safeguards and higher pricing for high‑risk uses, while enabling more flexible arrangements for lower risk applications that may align closely with research or development goals.

Tiered pricing in practice

Tiered pricing provides a practical way to reflect this nuance. Rather than a single headline rate, it combines three considerations.

First, the use case. Perpetual foundational model training usually sits in a higher tier than time‑limited grounding or fine‑tuning. This is because it has long‑term implications for content control, reputation and market substitution.

Second, the organization. Publishers often differentiate between large commercial AI developers, academic or mission‑led institutions, and early‑stage or under‑resourced initiatives. As a result, the same content may be priced differently depending on scale and intended downstream impact.

Third, the content. Books, journals, factsheets, databases and learning materials all vary in how they are produced and used. Some materials are short but highly structured and costly to maintain. This makes them valuable for AI use.

Linking prices to familiar publishing formats, such as those used on articles, books, datasets or institutional licences helps ensure content is valued for its quality and cost to produce, not just how long it is.

Mission‑led publishing in an AI world

For non‑profit publishers like CABI, AI licensing decisions sit within a broader set of responsibilities. There is a need to protect authors’ work and sustain publishing operations, while also ensuring essential knowledge remains accessible to those who depend on it.

Tiered pricing helps balance these objectives. It can incorporate discounts for academic or Global South use, support fixed‑budget or freemium models, and allow open‑access content to be included alongside licensed material under clear conditions.

Importantly, open access does not mean zero value. Openly licensed material still reflects significant investment in research, editorial oversight, translation and dissemination. They also contribute to responsible AI systems. In particular, where attribution, metadata and structured vocabularies are involved.

Pricing as part of responsible AI governance

Pricing, combined with clear licensing terms, transparency requirements and shared expectations around attribution and use, forms part of a broader governance framework.

Well‑designed licences can protect the rights of authors and publisher brands, and encourage AI systems to be grounded in reliable, domain‑specific knowledge.

As AI improves, demand is already shifting away from large volumes of web data to smaller, higher‑quality collection of texts used to train and fine tune models. This is precisely the content that specialist publishers curate.

CABI’s approach

Through the GAIA project, CABI has taken a deliberate and pragmatic approach. This includes developing market‑informed pricing guidance, a model content licence for AI use and approving priority subsets of content for responsible AI applications. Alongside this, assets such as the multilingual CABI Thesaurus knowledge graph* continue to play a key role in helping AI systems navigate complex agricultural and environmental concepts.

Together, these steps reflect a clear position. AI can strengthen, rather than undermine, trusted knowledge. However, this is only if the systems that use it respect its value, provenance and purpose.

For CABI, the goal is not to resist technological change, but to shape it. By engaging thoughtfully with AI through pricing, licensing and governance, we aim to ensure expert knowledge continues to be used ethically, transparently and responsibly in the next generation of AI tools.

---

*The multilingual CABI Thesaurus knowledge graph is a structured, concept‑based resource. It connects agriculture, environment, and life sciences terms across multiple languages to support consistent understanding, discovery, and data integration.

Further information

Featured image credit: Sorbetto (iStock)

Learn more about the ‘Generative AI for Agriculture’ (GAIA) project: access the pricing guidance and model content licence at https://www.cabi.org/projects/generative-ai-for-agriculture-advisory/.

Access the first subsets of CABI content approved for responsible AI use at: https://huggingface.co/datasets/CABInternational/ 

Related projects

‘Enabling FAIR data sharing and responsible data use’ project.

Related blogs

‘AI, copyright, and content licensing in digital agriculture’

‘Lessons in AI Governance from the GAIA Project’