Who owns the words and images AI creates for me?

You have an AI-drafted brochure, a set of product descriptions and a logo concept from an image tool, and a nagging question before any of it goes to print: is this mine? Can the AI company claim it? Can a competitor lift it? Could someone sue you for using it?

Untangle it into the three questions hiding in there, because they have different answers.

Can I use it commercially? Almost certainly yes

The major tools settle this by contract. OpenAI's terms assign you their rights in the output; Anthropic and Google take the same broad position for ordinary business use. You can publish it, sell with it, put it in client work. Two checks worth thirty seconds: image and design tools vary more than text tools (free tiers sometimes restrict commercial use or require attribution, so read the terms of the specific tool and plan), and output is not exclusive: if a competitor writes a similar prompt, they may get similar words, and neither of you can complain.

Do I own the copyright? Genuinely unsettled

This is where the lawyers earn their fees. UK law is unusual: the Copyright, Designs and Patents Act has a provision for "computer-generated works", giving authorship to the person who made the arrangements for the work to be created. On paper, that looks helpful. In practice, nobody yet knows how it applies to generative AI, courts have not settled whether a prompt counts as "making the arrangements", or whether purely AI-generated work gets copyright protection at all, and the government has been consulting on the whole area.

The practical consequence: the more human input, the stronger your position. Content you briefed in detail, edited, restructured and merged with your own knowledge is on solid ground as your copyright work. A raw, untouched generation is the weak case. So keep records of prompts and edits for anything valuable, and treat heavy editing not as cosmetic but as the thing that makes the asset properly yours. Since unedited AI output is usually mediocre anyway, the commercially sensible workflow and the legally sensible one happen to be the same workflow.

Could the output infringe someone else's rights?

Rarely, but not never. Models can reproduce protected material, an image steeped in a famous style, text close to its source. The risk concentrates exactly where you should not be using raw AI output anyway: logos, brand assets, anything you need to be distinctive and defensible. A logo is a trademark question as much as a copyright one, and "an image tool made it, I cannot register it confidently, and someone else can generate something similar tomorrow" is a terrible foundation for a brand. Pay a designer for the crown jewels; let AI do the volume work around them.

My own rule across everything I publish: AI does drafts, I do not ship anything I have not reshaped, and nothing brand-critical is raw machine output. Day to day, ownership worries should not stop you using these tools, the usage rights are clear and the sky is not falling. Just hold the one honest caveat: until the law catches up, "can I use it" is a yes, and "can I stop others" is a maybe, so build your moat out of the things AI cannot generate for your competitor: your data, your reputation and your name on the work.