Generative AI has developed at a remarkable pace, moving in a relatively short space of time from something experimental to something that is now being used across a wide range of business functions, from marketing and customer engagement through to content creation and internal operations.
As that shift continues, the nature of the conversation is beginning to change. It is no longer simply about what AI can do, but increasingly about how it works in practice and, more importantly, what it is built on.
In the creative industries, that question carries particular weight. Sectors such as publishing, journalism and the visual arts are rooted in the creation and curation of original work, so as generative AI systems begin to draw more heavily on that content, there is understandably more focus on how it is being used and on the wider questions that come with that.
At a basic level, the principle is relatively straightforward. If creative content is being used, then that use needs to be properly licensed, whether that happens through direct agreements between rights holders and developers or through more collective approaches.
That should not be seen as a constraint on innovation. If anything, it creates the conditions for innovation to develop with greater certainty, and in a way that is more likely to be sustainable over time.
Much of the attention around AI has, quite rightly, focused on the technical side, models, compute and performance. But it can sometimes overlook a more fundamental point, which is that the outputs these systems produce are shaped by the content on which they are trained.
Where that content is strong, diverse and reliable, the results tend to follow. Where it is not, the limitations become apparent relatively quickly, whether through inaccuracies, gaps in representation or a lack of consistency.
Alongside this, there is a growing focus on provenance. Organisations are not only asking whether AI systems perform effectively, but also where the underlying data comes from, how it has been sourced and whether it can be relied upon from both a practical and reputational perspective.
This is partly influenced by the evolving regulatory environment, but it is also being driven by the needs of businesses themselves. As AI becomes more embedded in day-to-day operations, there is a clear expectation that it should be built on foundations that are transparent, auditable and responsible.
In that context, clarity becomes more valuable than unrestricted access. Organisations want to understand what they can use, how they can use it and what the implications of doing so might be, particularly as they look to scale their use of AI with confidence.
The difficulty, of course, lies in the scale at which generative AI operates. These systems, such as large language models, rely on very large volumes of content, often drawn from a wide range of sources, and attempting to manage access and permission to use that content entirely through individual agreements is, in many cases, neither practical nor efficient.
This is where more collective approaches start to come into the conversation. Within the publishing sector and across the wider creative industries, there is increasing discussion around how collective licensing models, which are well established in other contexts, might be applied to the use of content in AI models. In practice, we are already beginning to see this take shape, including through the development of collective licences that publishers are now starting to opt into, offering a more structured route to responsible access to content at scale.
Such approaches offer a structured way of enabling access at scale, while maintaining clarity and consistency for both those using the content and those who have created it. They provide a mechanism through which large and diverse bodies of work can be made available in a way that is manageable and grounded in existing frameworks.
Importantly, this is not about replacing direct commercial relationships, which will continue to play an important role. Rather, it reflects a recognition that, in a landscape defined by scale, there is a need for complementary mechanisms that can operate effectively alongside them.
One of the consequences of not addressing this is the risk of narrowing participation. If access is limited to those organisations with the scale or leverage to negotiate directly, a significant proportion of the market is left out, including smaller publishers, independent creators and those operating within the long tail.
That has implications not just for fairness, but also for the quality and diversity of the content available to AI systems. A broader and more representative range of inputs tends to lead to better and more balanced outputs, which benefits the ecosystem as a whole.
At the same time, there is a noticeable shift in how content itself is being viewed. It is no longer simply something that sits behind AI systems, but something that actively shapes what those systems are able to do.
Technology enables the process, but the content itself determines the outcome.
That is why ideas such as transparency, lawful access and fair remuneration are becoming more prominent. Not as abstract principles, but as critical, practical considerations that underpin the development of AI in a way that is sustainable over time.
It also highlights how closely connected the different parts of the ecosystem are. Publishers, policymakers and AI developers are often characterised as having competing interests, yet in practice their roles are closely linked.
AI depends on access to high-quality content in order to function effectively. Creative industries benefit from technological developments that enable new forms of distribution and engagement. Policymakers are increasingly focused on establishing frameworks that support both.
The challenge, therefore, is not to resolve a simple tension, but to find approaches that allow these interests to coexist and reinforce one another.
For businesses, these questions are becoming more immediate. Decisions around the adoption of AI are no longer based solely on capability, but also on how systems are built, what they rely on and whether they can be deployed with confidence at scale.
Issues such as data sourcing, governance and accountability are moving from the margins to the centre of these discussions, reflecting their importance in shaping both risk and opportunity.
We are still in the early stages of working through these questions, but the direction of travel is clear. As AI becomes more embedded in business and society, expectations around how it is built will continue to rise.
Ensuring that the systems we develop are supported by responsible and sustainable approaches to content access is therefore not a peripheral concern, but a central one.
Because ultimately, the future of AI will not be determined solely by advances in technology, but by the strength and integrity of the foundations on which it is built.
