This article summarises publicly available guidance from regulators and official sources. It is general educational information only and does not constitute legal or professional advice. Requirements vary by jurisdiction. Consult your regional authority or a qualified professional for advice specific to your situation.
The UK government spent more than a year weighing whether AI companies should be able to train on copyrighted content by default, with creators having to opt out. On 18 March 2026 it published its answer: that plan is shelved, at least for now, and the existing, narrower copyright position stays in place.
In short: the UK is not introducing a broad text-and-data-mining exception with an opt-out for commercial AI training. The narrower exception that already existed, covering non-commercial research only, remains the current law. There is no near-term legislative change, and licensing between AI developers and content owners is being left to develop on an industry-led basis.
What the government was actually deciding
Since December 2024, the UK government ran a public consultation on how copyright law should treat AI training data, drawing 11,520 responses, the majority from the creative industries. Its findings and conclusions are set out in the official Report on Copyright and Artificial Intelligence, published 18 March 2026. The government's own previously preferred option was a broad text-and-data-mining exception that would let AI developers train on copyrighted material by default, with rights holders able to opt out if they didn't want their work used that way.
The owner of a 12-person UK marketing agency, using AI writing and image tools to draft client campaigns, had been assuming that opt-out model was coming, and that it might eventually mean more clarity, and more risk, around the AI tools her team already relied on. After the March 2026 report, the picture is different: the opt-out plan was dropped, and the practical rules her agency operates under haven't actually changed.
Why the government changed course
The government's own report points to concerns raised by the creative industries as the deciding factor: that a broad opt-out exception would undermine the value of copyrighted work, and that a workable opt-out mechanism doesn't actually exist yet. Robust technical standards for creators to signal and enforce an opt-out at scale weren't judged mature enough to build a legal framework around.
Rather than replace the opt-out plan with a different option, the government's report states there is currently no preferred legislative option at all. Instead, it's monitoring how licensing arrangements between AI developers and content owners develop on their own, and watching how other jurisdictions handle the same question, before considering any further legislative change.
What actually applies right now
The existing UK text-and-data-mining exception, which only permits copying copyrighted material for non-commercial research purposes, remains the law. A commercial AI developer training a model for a paid product does not fall within that exception. In practice, this means the legal position for commercial AI training in the UK hasn't shifted: it still generally requires a licence from, or permission from, the rights holder, exactly as it did before this year-long process began.
What this means for your business
If your business uses AI writing, image, or video tools, the tools themselves were trained under whatever licensing or legal basis the vendor already relies on, and that basis hasn't changed as a result of this report. It's still worth checking a vendor's own terms for what they say about the data used to train their models, particularly for any tool that generates content close to a specific brand, style, or named individual, since copyright risk in the output itself is a separate question from training-data legality.
If your business publishes content online, that content remains covered by ordinary UK copyright protection. The absence of a broad opt-out exception means there is currently no new legal mechanism that specifically permits an AI company to scrape and train on your published material without a licence, beyond the narrow non-commercial-research exception that already existed.
What to check in an AI vendor's terms
Since the underlying legal position hasn't shifted, the practical work for a business is unchanged too. When adopting a new AI writing, image, or video tool, look at the vendor's terms of service and any published model card for three things: whether the vendor states what data the model was trained on (even a general category, such as licensed stock libraries versus general web content); whether the vendor indemnifies customers against a copyright claim arising from generated output; and whether the tool's output could plausibly reproduce a specific artist's style, a named individual's likeness, or substantial verbatim text from a real source, since that's a separate infringement risk from the training-data question.
None of this requires a formal legal opinion for most everyday use. A UK business using an AI tool to draft marketing copy, summarise documents, or generate stock-style images for internal use carries relatively low risk under the current framework. The risk increases for output that's published at scale, sold as a standalone product, or generated in a style closely associated with a specific, identifiable creator.
Methodology (Real-World, Verified)
We test AI tools against real SMB workflows: the tasks a 20-person business actually uses AI for, not enterprise demos. Pricing is verified at the vendor's published rates, with local-currency conversions noted where relevant. Compliance notes reference the legislation and regulatory guidance relevant to each article's region. Every tool is judged on one question: could a business with no dedicated IT department actually pick this up and use it on Monday morning.
Related reading: our AI governance by region.
Did the UK introduce an opt-out for AI training on copyrighted content?
No. The government's March 2026 report abandoned its previously preferred option of a broad text-and-data-mining exception with an opt-out for commercial AI training. There is currently no preferred legislative option.
Can AI companies legally train on my content without permission in the UK?
For commercial AI training, generally no. The UK's existing text-and-data-mining exception only covers non-commercial research purposes. Commercial AI training on copyrighted material typically still requires a licence from the rights holder under current UK law.
Does this change which AI tools are safe to use in my business?
Not directly. The legal basis your AI vendors already rely on for training their models hasn't changed as a result of this report. It's still good practice to check a vendor's terms for how their models were trained, particularly for tools that generate brand- or style-specific content.
Is the UK likely to revisit this decision?
The government's report frames this as an ongoing, monitored position rather than a final one. It's watching how industry-led licensing arrangements develop and how other jurisdictions handle the same question before considering any further legislative change.
The information in this article is general in nature. It reflects a summary of publicly available guidance and does not constitute legal, privacy, or professional advice. Your obligations will depend on your specific situation, jurisdiction, and business circumstances. Do not rely on this article as a substitute for qualified legal or professional advice.
This is one piece of the UK's wider AI regulatory picture. See how it fits alongside data protection, sector regulators, and the UK's principles-based approach.
Read the UK AI Governance Guide