A professional services firm in Glasgow had been using AI writing tools for two years when a freelance copywriter sent a polite but pointed question: was any of the content delivered to clients produced with AI trained on licensed material? The answer was: we don’t know. That’s the honest answer for many owner-managed businesses in the UK right now.
In March 2026, the UK government published its long-awaited report on copyright and artificial intelligence. The headline conclusion: current law stays unchanged. There is no new exception allowing businesses or AI developers to use copyright-protected content for commercial AI training. The government stepped back from an earlier proposal, committed to more evidence-gathering, and left the legal landscape for AI users exactly as it was. That ambiguity has practical consequences.
What is the UK government’s current position on AI copyright?
The UK government published its Report on Copyright and Artificial Intelligence in March 2026, under the Data (Use and Access) Act 2025. The headline finding: current copyright law remains unchanged. There is no new commercial exception allowing AI developers or businesses to train models on protected content without permission. The government favoured licensing over new exceptions and committed to more evidence-gathering before legislating.
The only existing text and data mining exception in UK law is section 29A of the Copyright, Designs and Patents Act 1988, and it has always been limited to non-commercial research. It was never designed for commercial AI training.
The government had floated a broader approach ahead of this report. Its 2024-25 consultation proposed a commercial text and data mining exception with an opt-out for rights holders. After more than 11,000 consultation responses, many from creative industries opposing any broad exception, the government dropped the opt-out proposal. Legal commentators at Simmons & Simmons described the March 2026 report as “a masterclass in can-kicking.” The government maintains the status quo and commits to monitoring litigation and overseas developments before moving further.
Why does this matter for your business?
Many owner-managed businesses are using AI tools rather than building their own models, which makes copyright risk feel like someone else’s problem. But the March 2026 report leaves three questions genuinely open: who owns the AI-assisted content your business produces, how reliable are the IP indemnities in your tool contracts, and what exposure do you carry if training data was used without permission.
The ownership question comes from a specific provision in UK law: section 9(3) of the Copyright, Designs and Patents Act 1988. This grants copyright protection to computer-generated works where there is no human author. The government acknowledged “minimal evidence” that this protection is being used or producing significant economic effect, and has flagged it for possible repeal. Until that decision is made, the copyright status of AI-generated content sits in an uncertain position.
For businesses that deliver content to clients, or that use AI to produce internal documentation, marketing copy, or reports, the practical risk has two sides. The first is whether the output you rely on is actually protected as your intellectual property. The second is whether your AI vendor’s training data creates any downstream exposure for you, and whether their contract terms adequately address it.
Where will you actually encounter this risk?
For owner-managed businesses, copyright exposure shows up in three distinct places: who owns the content an AI tool generates for you, whether the tool was trained lawfully on third-party material, and what happens when you use customer data to fine-tune a model. Each involves different law and different risk, and none are clearly resolved by current UK legislation.
The content ownership question is the most immediate. Under current law, works with significant human creative input are protected as the author’s copyright. Purely machine-generated works fall under the computer-generated works provision, which could be repealed. Legal advisers consistently recommend documenting the human creative contributions to every piece of AI-assisted work, and maximising that input where possible.
The training data question applies even if you never train a model yourself. If the AI tool you use was trained on unlicensed content, your outputs could carry that risk. The CMA has examined these dynamics in its work on foundation models, highlighting how contractual imbalances can shift risk to end users without making that clear in the terms of service. Read the IP indemnity clauses in any AI contract before relying on them.
The personal data question applies whenever you consider feeding client or customer information into an AI system for fine-tuning or analysis. The ICO has made clear that using personal data in AI processes requires a lawful basis under UK GDPR, and a Data Protection Impact Assessment where the processing is high-risk.
When do you need to act, and when can you wait?
The government’s “wait and see” position gives some breathing room, but it doesn’t remove existing risk. The Copyright, Designs and Patents Act 1988 still applies, and there is no statutory permission for commercial AI training on third-party content. What the March 2026 report removes is the immediate threat of sudden legislative change, giving you time to put basic protections in place without scrambling.
What can reasonably wait: building compliance systems designed around legislation that hasn’t been written yet. The government has said it prefers licensing and market solutions over immediate reform. The House of Lords Digital and Communications Committee inquiry into copyright and AI will produce recommendations that shape the next phase of policy. Further change is coming, but not overnight.
What deserves attention now: the contracts and processes already in use. Three specific checks are worth making this quarter. Read the IP and indemnity clauses in your AI tool agreements, including what they say about training data and copyright in outputs. Establish a practice of maximising human creative input in AI-assisted work and keeping a record of that contribution. And if you’re using customer or client data in any AI process, confirm you have a documented lawful basis for that processing.
What else should you be watching?
Three adjacent areas are worth watching. The government plans a summer 2026 consultation on digital replicas and personality rights, which will matter for any business using AI to generate synthetic voices or likenesses. EU AI Act obligations on training data transparency will reach UK businesses through their AI vendors. And the ICO’s guidance on AI and data protection is already live, with real compliance expectations attached.
The digital replica consultation covers AI-generated likenesses, voices, and avatars of real individuals. If your business uses synthetic presenters, voice clones, or AI-generated imagery involving identifiable people, new consent and licensing requirements could follow. The government has cited both the risks of deepfakes and the need to support legitimate uses.
The EU AI Act, which has been in force since 2024, requires general-purpose AI model providers to publish summaries of the training data they used. UK businesses selling AI services into EU markets, or using EU-market AI vendors, will feel this through their supply chains. UK law does not mirror the requirement, but it remains on the watchlist.
The ICO’s guidance on AI and data protection has been live since the early phase of the generative AI rollout. If your team is using AI tools that process information about individuals, including customer contact details, employee records, or call transcripts, the ICO expects you to have a lawful basis for that processing and appropriate safeguards in place.
The common thread across all of this: talk to your contracts before you talk to your lawyers. The terms you agreed to when you signed up for AI tools carry more of your actual risk than the legislation still being drafted.
