Freelance and AI content: who owns the rights to generated work

A designer, copywriter, or illustrator increasingly delivers a result where part of the text, layout, or image was made by a neural network. The client expects “rights ready, as in an ordinary statement of work,” and the contractor expects payment without a claim six months later. Law 243-FZ and the practice of civil-law and services contracts require an explicit agreement: who is the author of the result, who owns the exclusive rights, whether AI labeling is required, and who is liable if the generation infringes someone else’s rights or a platform policy.

Why “I just delivered the layout” no longer works

A classic services contract is often silent on tools: it assumes a person made the result. With generative models, three separate questions appear: (1) what exactly is transferred — source files, prompts, the final file; (2) whether AI may be used, and in what volume; (3) who bears the risk of IP and labeling claims. Without these blocks, the dispute comes down to “we never agreed that.”

The general context on 243-FZ and contracts with contractors is in AI law and contracts in 2026. Below, the focus is the freelance setup: one contractor, one client, content made to order.

What to write into the contract

1. Permission to use AI, and the scope of use

• Whether generative models may be used (yes / no / drafts only)
• Which stages: idea, text, visual, code, voiceover
• A duty to tell the client that AI was used, before delivery
• A ban on uploading the client’s confidential data to public services without consent

2. Rights to the result and to the “raw material”

• The moment exclusive rights pass (payment / certificate / delivery of files)
• Whether rights to prompts, intermediate generations, and variations pass as well
• Whether the contractor keeps a portfolio right, with NDA reservations
• Licenses for third-party stock and models — who pays for them, and in whose name

3. Labeling and liability

• Whether AI-content labeling is required for the client’s platforms (advertising, a marketplace, media)
• Who prepares the labeling wording, and where it is placed
• The contractor’s warranties: the result does not infringe others’ rights “within a reasonable check”
• A liability cap and the claims procedure if there is an IP complaint

Typical imbalances between the parties

1. The client demands “all rights worldwide” and a ban on AI — at the price of template text
2. The contractor stays silent about the neural network, and it later turns out the service license does not allow a commercial transfer to the client
3. The contract has no ban on leaking the brief and personal data into a public chatbot
4. Labeling is “hung” only on the freelancer, although the client’s marketing team does the publication
5. Edits are accepted in a messenger with no record of whether the delivery was “with AI” or “without”

A minimum block for a contract annex

Even a short annex removes half the disputes. It is worth closing these points explicitly:

• A list of permitted AI services, or a criterion of “commercial license only”
• A ban on entering trade secrets and personal data into third-party models without written consent
• The delivery format: source files + final + a list of tools
• The labeling rule when the client publishes
• The rework procedure if a platform rejects the content because of an AI policy

A draft services or civil-law contract can be assembled from the brief: draft a contract.

A 10-minute checklist

Related reading

• AI law and contracts in 2026
• Contracts for freelancers
• How to draft a contract with a self-employed person
• Hidden clauses in a contract

Short conclusion

A freelancer’s rights to AI content do not “arise on their own”: they have to be allocated in the contract together with the rules on labeling, confidentiality, and liability. The earlier the parties record the tools and the scope of the rights transfer, the smaller the chance that delivery falls through because of a platform policy or a third-party claim.