Copyright and licensing

Copyright is the legal right that gives creators, or whoever holds the rights, control over copying, adapting and sharing original work. A licence is permission from the rights holder to use that work under stated conditions. Together they decide what an organisation can publish, quote, adapt, translate and reuse, and what others may do with the organisation’s own content.

Content teams meet them everywhere: stock photographs, commissioned copy, charts, quotations, interview transcripts, open data and a paragraph copied into a slide deck because it was “already on the internet”. Being available isn’t the same as being permitted. Details vary by country, but the basic pattern is broadly shared: copyright arises automatically, lasts for decades and covers the expression of an idea rather than the idea itself.

Why it matters for content

Ownership should be settled before publication, especially for commissioned work and material made with partners. A contract may pay for an asset without transferring copyright. It should also cover practical uses, such as whether a commissioned illustration may be translated, adapted or used in another publication. Stock licences often limit where and how long an image can appear.

Creative Commons licences give a standard way to grant permission in advance. All of them require attribution. Some add conditions: NonCommercial restricts commercial use, NoDerivatives forbids sharing adapted versions and ShareAlike requires adaptations to carry the same licence. These are different conditions and shouldn’t be treated as interchangeable.

Licensing can also be a deliberate choice about your own content. An organisation may want people to quote its guidance, adapt training materials or translate a public resource. The Content Lexicon, for example, can be quoted freely as long as the quote links back. A clear statement like that removes unnecessary permission requests and protects the condition that matters.

How AI changes the picture

Generative AI has made the boundary between reading, copying and training hotly contested. Different jurisdictions are reaching different answers. In the US, dozens of lawsuits turn on whether training counts as fair use. The UK government’s 2026 report describes a developing licensing market, unresolved evidence and several possible policy approaches.

That leaves content owners with a newer question: whether to let AI companies train on their content, and how to say so. The tools work at different levels. Robots.txt rules can ask named crawlers to stay away, but only crawlers that choose to respect them. Licensing deals set terms with specific AI companies. Machine-readable licensing standards, such as RSL, let a publisher state terms for AI use alongside its content, from free with attribution to paid. Proposals such as llms.txt help AI tools find content rather than set permissions. None of these is a universal switch, so decide the position first and choose the signals that express it.

Creative Commons warns that choosing a more restrictive CC licence may block useful human reuse without reliably stopping AI training. For material with real risk, take legal advice rather than relying on folklore passed around a group chat.