AI and Intellectual Property: A Practical Guide for UK and Irish Businesses

Businesses across the UK and Ireland are using AI tools every day — to write marketing copy, generate product images, draft software, and build systems that learn from data. Most have no formal policy on who owns what comes out the other end. That gap is not theoretical. Copyright tribunals, patent offices, and courts in the UK, EU, and US are actively drawing boundaries, and the positions they’re taking will affect every business using generative AI commercially.

This guide explains where the law currently stands on AI and intellectual property, what the recent case law means for your business, and the specific steps you can take now to protect the work your teams produce with AI tools.

What AI and Intellectual Property Actually Mean for Your Business

AI and intellectual property is not purely an academic concern for legal departments. It touches any business that uses AI tools to create content, code, or products — which, at this point, is most of them.

The core tension sits in two places. First, AI models are trained on existing data, much of which is protected by copyright. Second, the outputs AI produces may not be eligible for copyright protection at all, because most jurisdictions require a human author for copyright to exist. Both problems carry real commercial risk.

The Input Problem: Training Data and Copyright

When a business trains an AI model on data it does not own, or uses a third-party model trained on scraped web content, it may be receiving outputs that carry infringement risk. This is not hypothetical. Getty Images filed suit in the US in 2023 against a generative AI company over the use of its image library for training. Several similar cases are ongoing in the US and UK courts.

The UK’s position is more permissive than the EU’s. Under the Copyright, Designs and Patents Act 1988, there is a text and data mining (TDM) exception for research, though its commercial scope is contested. The EU AI Act and the EU Copyright Directive take a stricter approach: AI providers must disclose the training data used, and rights holders can opt out of having their content used for commercial training. For businesses operating in Ireland or Northern Ireland, this matters directly.

The Output Problem: Who Owns What AI Creates

UK copyright law requires a human author. Under the Copyright, Designs and Patents Act 1988, computer-generated works can attract copyright, but the “author” is defined as the person who makes the arrangements necessary for the work to be created. What “necessary arrangements” mean in practice is being contested in courts right now.

The US Copyright Office has been clearer: it will not register purely AI-generated works. The EU position is similar. All three jurisdictions accept that a work produced with meaningful human creative input can be protected. The question is how much human input is enough.

The Human-in-the-Loop Requirement

If there is one concept that should drive how your business handles AI and intellectual property, it is this: document the human creative contribution at every stage of production.

What “Human Authorship” Means in Practice

Courts and IP offices are not looking for token involvement. A single prompt typed into ChatGPT, with no further creative input, is unlikely to meet the threshold for copyright in the UK or the US. What they are looking at is whether a person made substantive creative choices: selecting, arranging, editing, directing, and refining the output.

The UK Intellectual Property Office has indicated that the more a human shapes the final work — through iterative prompting, editing, combination with original elements, or creative direction — the more likely it is to attract protection. There is no defined numerical threshold. The test is qualitative.

Building a Human-in-the-Loop Workflow

For businesses using AI tools to produce commercial content or code, the practical response is a documentation-first approach:

Keep prompt logs that show the instructions given and the creative direction applied. Save version histories that demonstrate the editing and selection process after initial AI output. Record who made which decisions and when. If the work is collaborative between a human and an AI tool, make sure the human’s contribution is clearly visible in the production record.

This is particularly important for software businesses. AI-generated code may carry infringement risk from training data (especially if the model was trained on open-source code with licence conditions) and may not attract copyright protection on its own. A development team that uses GitHub Copilot or similar tools without documenting its own additions and modifications is building on uncertain ground.

Key UK Case Law: Thaler v Comptroller-General

The UK Supreme Court ruled in 2023 that an AI system cannot be named as an inventor on a UK patent. The case, brought by Dr Stephen Thaler regarding his DABUS AI system, established that the Patents Act 1977 requires a human inventor. The patent application failed not because the invention was invalid, but because no human inventor could be identified.

The implication for businesses developing AI-assisted products is direct: if you want patent protection, a human must be the named inventor, and you need to be able to demonstrate that human’s genuine inventive contribution.

The UK, Ireland, and Northern Ireland Regulatory Divergence

Most published guidance on AI and intellectual property treats “the UK” and “the EU” as two separate, monolithic systems. For businesses based in Northern Ireland, or those operating across the island of Ireland, the picture is more complicated.

The EU AI Act and What It Means for Irish Operations

The EU AI Act came into force in August 2024 and is being phased in through 2025 and 2026. For businesses in the Republic of Ireland, it applies directly. Among its IP-relevant provisions, the EU Copyright Directive requires providers of general-purpose AI models to publish summaries of the training data used, and it allows rights holders to opt out of having their content used for commercial AI training.

For Irish businesses using third-party AI tools, this creates a due diligence requirement: you should understand what your AI provider has disclosed about its training data, and whether that disclosure meets EU requirements. For businesses developing their own AI models, compliance with the training data transparency requirements is not optional.

The UK’s Post-Brexit Stance

The UK has taken a deliberately lighter-touch approach to AI regulation, positioning itself as a pro-innovation jurisdiction. The UK government shelved proposals to mandate transparency on AI training data after industry pushback in 2023. The result is that UK businesses operating purely in the UK market face fewer formal disclosure obligations on training data than their counterparts in the EU.

That gap has implications for businesses on both sides of the Irish border. A Northern Ireland business trading into the Republic may need to meet EU AI Act obligations for the Irish-market part of its operation, even if its Northern Ireland operations are not directly subject to EU law. The Windsor Framework does not directly govern IP law, but in practice, many NI businesses need to meet both sets of expectations.

ProfileTree works with SMEs across Northern Ireland, Ireland, and the UK on digital strategy and AI implementation, and this regulatory divergence is one of the most practically complex issues we see for cross-border businesses. Getting clarity on which regime applies to which part of your operation is a genuine first step, not an afterthought.

A Quick Decision Framework for NI Businesses

If your business is based in Northern Ireland and sells goods or services into the Republic of Ireland or the wider EU, you should assume EU AI Act obligations apply to the EU-facing part of your operation. If you use AI tools to generate content or products sold into that market, the EU AI Act’s transparency requirements for training data are relevant. If you are purely UK-focused, the current UK regime is lighter, but that may change as the UK’s AI governance framework develops.

Protecting Your Business: Practical IP Strategies

Understanding the law is useful. Protecting yourself requires action. The following approaches are relevant regardless of whether you are a software company, a creative agency, a manufacturer, or a service business using AI tools for marketing and operations.

Conduct an AI Usage Audit

Before you can protect anything, you need to know what you have. An AI usage audit maps every point in your business where AI tools are being used to create outputs with potential commercial value.

For each tool, record: what it is used for, what inputs go in, what outputs come out, who has creative control over those outputs, and what the tool’s terms of service say about IP ownership of outputs. Most commercial AI tools (including ChatGPT, Midjourney, and GitHub Copilot) include clauses on ownership of their outputs in their terms of service. Some assign ownership to the user; others are ambiguous. You need to know what you have agreed to.

Establish Contractual Safeguards

If you are producing AI-assisted work for clients, your contracts need to explicitly address IP ownership of the output. Do not rely on standard terms that were written before generative AI existed. Key clauses to include or review:

Ownership of AI-assisted outputs should be clearly assigned. Indemnity provisions should specify who bears liability if an AI-generated output is found to infringe a third party’s rights. If a client brief requires you to use specific AI tools, the contract should specify whose terms of service govern the output. For software development, provisions on AI-generated code should address licence risk from training data.

Consider Trade Secrets Over Patents for AI Workflows

For many businesses, the most valuable IP asset in an AI workflow is not a patentable invention but a proprietary process: the prompting approach, the fine-tuning method, the training dataset, or the combination of tools that produces a distinctive result. Trade secret protection does not require registration or disclosure of the underlying method. It does require that you treat the information as confidential and have reasonable measures in place to protect it.

In practical terms: document your AI workflows, restrict access to proprietary processes, include confidentiality provisions in employment and contractor agreements, and keep clear records of what makes your approach different from publicly available methods.

Use AI Tools Safely in Marketing and Content

For creative agencies and marketing teams, the practical question is: can we use AI-generated images and copy in client campaigns without IP risk? The answer is: it depends on the tool, the terms of service, and the extent of human creative input in the final output.

For image generation tools, check whether the provider offers commercial licences and indemnification. For written content, the risk is less about copyright (since AI-generated text is unlikely to attract copyright for a third party) and more about ensuring your own ability to protect the content you produce. The more human editorial work goes into the output, the stronger your position.

Our AI implementation services for SMEs include practical guidance on building AI workflows that protect your commercial interests, and our AI prompts for business resources cover how to structure prompting for quality and accountability.

Patents, Trademarks, and Trade Secrets in the AI Context

The three main IP categories each interact with AI differently.

Patents and AI-Generated Inventions

Following the Thaler ruling, UK patent applications for AI-assisted inventions must name a human inventor. The EPO (European Patent Office) takes the same position. In practice, this means that if your development team uses AI tools in the invention process, you need to ensure that a human inventor’s genuine contribution is identifiable and documented. The AI’s role can be acknowledged as a tool, like software or laboratory equipment, but it cannot replace the human inventor.

For businesses in technology sectors, this is a real planning issue. If you are developing products with significant AI input during the design or engineering phase, early-stage legal advice from a patent attorney is worthwhile. The cost of proper inventorship documentation is minor compared to the cost of a patent that cannot be enforced.

Trademarks and AI Branding

A trademark protects a sign that distinguishes your goods or services from those of others: a name, a logo, a shape, sometimes a colour. The trademark system does not care whether the logo was designed by a human or generated by AI, as long as it is distinctive and not confusingly similar to existing marks. However, if your AI-generated brand assets are not protected by copyright (because there was insufficient human authorship), your protection rests entirely on the trademark registration. That makes registration more important, not less.

Trade Secrets: Often the Right Answer

For most SMEs, trade secret protection is more practical and immediately valuable than patent protection for AI-related IP. You do not need to register, you do not need to disclose the underlying method, and the protection does not expire. The conditions are that the information has commercial value, is not generally known, and you take reasonable steps to keep it confidential.

An AI training dataset compiled from your own business data, a proprietary prompting framework, or a fine-tuned model built on your operational knowledge can all potentially be protected as trade secrets, provided you treat them as such.

Practical IP Checklist for AI-Using Businesses

Before your next review cycle, work through these questions:

Have you reviewed the terms of service for every AI tool your team uses? Do your client contracts address IP ownership of AI-assisted outputs? Do you have a documentation process for human creative input in AI workflows? Have you identified which of your AI processes or datasets could be protected as trade secrets? If you are in Northern Ireland or Ireland, do you know which AI regulatory regime applies to which part of your operation? Have you taken legal advice on whether any AI-assisted products or inventions should be protected by a patent, and whether inventorship is correctly documented?

This is not an exhaustive legal audit, but it is a starting checklist. The businesses that act on these questions now will be in a stronger position as the law continues to develop.

As Ciaran Connolly, founder of ProfileTree, puts it: “We’re at a point where almost every SME is using AI in some part of their workflow. The ones who treat IP documentation as a parallel process to the work itself, rather than something they’ll deal with later, are the ones who’ll actually own what they’ve built.”

For businesses implementing AI and looking to protect their digital assets, our digital marketing training covers the responsible use of AI tools, and our broader work on ethics and legalities in digital marketing provides context on where compliance and commercial strategy intersect.

Conclusion

AI and intellectual property law are moving faster than most businesses can keep up with. The core principles are settled enough to act on now: document human creative input, audit the tools you use, update your contracts, and understand which regulatory regime applies to your operation. Businesses that build these habits into their AI workflows today will own what they create. Those who leave it until a dispute arises will find the answers are more expensive to obtain.

Contact ProfileTree to discuss how to implement AI tools safely and strategically across your business.

FAQs