Business analysis
AI and copyright: what should a small business organise now?
On 13 July, the European Commission published a study on a possible EU-level registry for rightsholders to express text-and-data-mining opt-outs. No registry has been adopted, but the study highlights a practical question: does a business know which content it owns, what it is licensed to use and what it gives to AI services?

14 July 2026 at 06:30
What happened?
The Commission study assesses the policy case and technical feasibility of a shared EU registry. Such a registry could in future help rightsholders state that protected material may not be used for commercial text and data mining. It is still a study: no new registry, deadline or new small-business obligation has been adopted.
Who should pay attention?
The issue matters to businesses that publish original photos, text, video, courses or product data online. It also matters when outside creators are used or material is uploaded to an AI service. An ordinary user of an off-the-shelf AI tool is generally not the provider of a general-purpose AI model, but the business still needs the right to use its inputs and terms suitable for customer or confidential data.
1. Make a lightweight content inventory
For important text, images, video, audio and datasets, record the creator, source, owner, licence and permitted uses. Separate in-house material, customer material, stock assets and AI-assisted content. A file being present in a company folder is not proof of permission.
2. Check contracts and AI settings
Agree in writing who owns material produced by a freelancer or agency and where it may be used. For AI services, check input retention, possible use for model improvement, user permissions and deletion. Do not upload customer data or confidential material merely because a tool is convenient.
3. Separate rights reservation from internal policy
Article 4 of the EU copyright directive permits text and data mining of lawfully accessible material under certain conditions when the rightsholder has not reserved the rights appropriately; for public online content this may mean machine-readable means. A single robots.txt line does not settle ownership, licences or every kind of use. First decide what the business wants to allow, restrict or license, and only then implement a technical signal.
4. Keep a trace for AI-assisted content
Record at least the service used, date, responsible person, key source material and the human review. This makes corrections, customer questions and later updates easier even if service terms or tools change.
A practical 30-minute start
Choose five business-critical content assets and note the owner, usage permission, storage location, AI use and responsible person for each. If something is unclear, do not guess: mark it for checking against a contract, service terms or professional advice.