Who Owns the Output of an AI Employee?

When your business uses a generative AI tool to draft a marketing email, write computer code, or design a new logo, the question immediately arises: who owns that completed work? The answer is not simple, as it depends on a combination of current intellectual property law, the specific tool's terms of service, and how the AI was used. This issue is particularly critical for UK businesses aiming to protect their digital assets and understand the boundaries of copyright in the age of automated services.

How Does UK Copyright Law Apply to AI-Generated Content?

In the United Kingdom, copyright law is governed by the Copyright, Designs and Patents Act 1988 (CDPA). Traditionally, copyright is granted to the human author who created the original work through the exercise of skill, labour, or judgement. This framework is designed to protect human creativity.

Generative AI, however, challenges this principle because the "author" is not a person, and the creation process involves automated algorithms rather than human creative choices in the same way.

Current UK copyright law is complex. It does contain provisions for computer-generated works that have no human author, but these are rarely tested and create significant ambiguity for modern generative AI. Under the CDPA, for a literary, dramatic, musical, or artistic work generated by a computer in circumstances such that there is no human author, the person by whom the arrangements necessary for the creation of the work are undertaken is deemed to be the author. For works created by modern generative AI, this could potentially be the person who writes the detailed prompt, but this is legally uncertain and depends heavily on the specific situation and the level of human involvement.

Who is the Human Author in AI Creations?

For most content created today, the primary issue is originality. Copyright protects original expressions, not ideas or automated outputs. When a user provides a very basic prompt, like "Write a formal email about a product launch," and the AI generates the entire text, the human's contribution of "skill, labour, or judgement" is minimal. The specific phrasing and structure of the email were chosen by the algorithm based on its statistical models, not by the user. In such scenarios, most legal experts in the UK would argue that no human copyright exists in the output.

If, however, a person writes an exceptionally detailed, complex prompt, and then iteratively refines the output, significantly editing and modifying the text, the resulting work is a collaboration. Copyright would then protect the specific human-edited version of the content, but not necessarily the raw AI output that served as the base. The legal "skill, labour, and judgement" are considered to come from the human’s editing, prompt-writing, and selection, not the automated system.

How Do AI Terms of Service Affect Ownership?

Even if copyright law is unclear, the contractual agreement you have with the AI service provider can determine ownership. When you sign up for and use a generative AI platform (like OpenAI's ChatGPT, Midjourney, or Anthropic's Claude), you agree to their Terms of Service (ToS).

You must carefully review these terms, as they are not uniform:

1. Full Rights Assignment: Some AI providers explicitly state in their terms that they assign all right, title, and interest in and to the output to the user who provided the prompt. For business users, this is the most desirable agreement.

2. Usage Licenses (Not Ownership): Other services might grant you a non-exclusive license to use the output for your business (even commercially) but retain ownership of the underlying generated data themselves. This is critical for assets like logos, as you might not be able to prevent others from using the same design.

3. Specific Restrictions: The terms may contain restrictions, for instance, prohibiting the creation of certain content or stating that rights are only granted after payment of specific fees.

The contract will generally override general copyright assumptions, so your primary protection comes from the service agreement itself.

Can You Protect the Content Your Business Generates with AI?

Given the legal uncertainty surrounding original AI output, UK businesses must focus on what is protectable.

• Human Modifications: If your marketing team uses AI to draft a 500-word article but a copywriter heavily edits, rewrites, and adds original human insight to 200 of those words, the copywriter's specific contributions are protected by copyright. The final article is a hybrid, and copyright covers the original human-created elements.

• Compilation Rights: If a business uses an AI-powered cloud service or a virtual desktop to generate 50 pieces of market analysis and a human selects, organises, and arranges these into a "Quarterly Market Report," that specific arrangement and selection can be protected as a database or compilation, even if individual AI-generated pieces within it have no separate copyright.

• Trademark for Brand Identity: For critical brand elements like a logo, trademark registration is often more valuable than copyright. If you generate a logo with AI and it meets the criteria for trademark registration, you can register it. A trademark does not require you to prove original authorship in the same creative sense that copyright does; it protects distinctiveness as a source indicator in trade. However, you still need to ensure you have the contractual rights from the AI provider to use the design this way.

The key message for non-technical users in the UK is that relying purely on the output of an AI system provides weak protection for your business assets. Ownership and protectability come from human interaction, editing, selection, and the service contract, not from the algorithm. To secure digital assets effectively, particularly when managing teams using remote tools, robust service agreements and clear human oversight are essential.