The UK government is drawing a firmer line around AI and copyright. A broad exception for text and data mining by AI systems was considered, but it is now off the table.
The decision matters for both AI providers and creative industries. It keeps licensing and existing exceptions at the center of the debate, rather than giving AI systems a wide new route to use protected material without permission.
No broad exception for AI text and data mining
Text and data mining is the process of extracting information from large amounts of text or data using AI and machine learning. The UK government had considered a broad copyright exception for this activity in relation to AI systems.
That approach will not move forward. Instead, the government has decided not to introduce broad copyright exceptions for text and data mining for AI systems.
This does not mean the government rejects AI. The source makes clear that the government recognizes potential benefits for creative industries, including increased productivity and efficiency. But it is also weighing those benefits against concerns from media and creative sectors.
Those concerns include the effect of AI-generated content on human creativity. In practical terms, the government is trying to avoid a framework where AI growth comes at the expense of the copyright system that is meant to reward creative investment.
AI output can still create copyright risk
The government’s position also reaches beyond training. According to the government, AI’s reproduction of copyrighted works violates copyright law unless it is permitted by a license or exception.
That statement puts pressure on AI systems that can reproduce protected content in their outputs. The source notes that it is not clear whether the government means exact copies only, or whether a close approximation of the original could also be enough.
This distinction is important. Exact reproduction is easier to identify and argue about. Close approximation would be harder to regulate, because it could create many more possible disputes over whether an AI-generated result is too close to protected material.
The New York Times lawsuit is one reason this issue has become sharper. The source says the New York Times was able to prove exact reproduction of training material for the OpenAI models. From the UK government’s standpoint, that kind of exact reproduction is likely to constitute copyright infringement.
OpenAI’s position raises the stakes
Before the New York Times lawsuit, OpenAI told the UK government that it was impossible to train foundational AI models without copyrighted data. That position points to the central tension in the debate.
On one side, AI developers argue that large-scale training depends on access to large bodies of material. On the other side, publishers, media companies and creative workers worry that their copyrighted works may be used or reproduced in ways that undermine the value of human-made content.
The source also notes OpenAI’s view that exact copies are a "rare bug" that can be fixed. If courts accept that view, the issue may be treated as a technical flaw that providers can reduce or correct.
But the source raises a different possibility. If courts conclude that outputting training material is part of the core function of AI models, providers could face much greater difficulty. In that case, they might not be able to guarantee that copyright infringements would not happen again and again.
A code of practice is the next step
The UK government intends to develop a code of practice for copyright and AI. The goal is to allow the AI and creative industries to grow in partnership while keeping the UK’s copyright framework focused on encouraging and rewarding investment in creativity.
The government is working closely with AI and creative industries stakeholders on that effort. The code of practice is expected to be published in early 2024.
For AI companies, the message is that copyright risk is not limited to lawsuits after publication. It can affect how models are trained, how outputs are controlled, and how providers explain the behavior of their systems.
For creative industries, the decision keeps copyright permission, licensing and exceptions at the center of the discussion. It also signals that the government sees value in AI, but not at the cost of removing broad protection for copyrighted works.
What this means for AI and creative work
The UK position does not settle every question. It leaves open whether close approximations of protected works should be treated like exact reproductions. It also leaves open how reliably AI providers can prevent protected training material from appearing in outputs.
Still, the direction is clear. The government is not offering a broad text and data mining exception for AI systems, and it views AI reproduction of copyrighted works as a legal problem unless a license or exception applies.
That makes copyright compliance a core issue for AI development, not a side concern. The emerging code of practice will be important because it is meant to define how AI and creative industries can work together without weakening the framework that supports investment in creativity.