Should Freelancers Tell Clients They Use AI in 2026?

The regulatory one is the EU AI Act. Its transparency rules under Article 50 apply from 2 August 2026. They require that synthetic audio, image, video, or text be marked in a machine-readable format and be detectable as artificially generated, and that people be told when content is an AI deepfake or when AI-generated text is published to inform the public on matters of public interest Source. Systems already on the market get a short grace period, until 2 December 2026, to meet the machine-readable marking part. If you serve EU clients or audiences, this is now a date in your calendar, not a think-piece.

The second trigger is the client's own AI policy. Many companies now restrict generative tools internally, often hard. Roughly 70% of organizations that block ChatGPT say they do it to protect confidential information, and names like Samsung, Amazon, and Verizon have all limited employee use after data scares Source. If your client bans these tools for their own staff, they almost certainly expect the same from the contractor they handed the same data to. Assuming the rule stops at the building's edge is how freelancers walk into trouble.

The third trigger is the platform you sell on. Fiverr's guidelines tell freelancers to disclose AI use when a buyer asks, and warn that selling a gig as custom written or hand-crafted while delivering undisclosed AI work breaks buyer trust and can violate the terms of service Source. The platform's framing puts a lot of the burden on the buyer to object first, which I think is backwards, but the practical takeaway is the same: if your marketing implies a human made every word, undisclosed AI is a problem.

What about the NDA you already signed?

The quiet risk is confidentiality. Most NDAs forbid sharing the client's confidential information with third parties. A public AI tool is a third party. Paste the brief, the source code, or the customer list into a consumer chatbot and you may have handed protected material to an outside service that can retain or train on it, which is the kind of disclosure your NDA was written to prevent.

This is not theoretical. Corporate legal teams started warning staff against pasting proprietary data into these tools precisely because public models can learn from what users type, and that input could resurface elsewhere Source. A signed NDA does not list every banned tool by name, but a court does not need a named list to find that you moved confidential data somewhere you should not have.

So there are two distinct questions, and freelancers collapse them at their peril. One is "did I use AI," which is about credit and process. The other is "did I feed the client's protected data into an outside system," which is about a possible contract breach. You can do the first cleanly and still get sued over the second. If you want the deeper version of how confidentiality and AI clauses interact, we cover it in anti-AI contract clauses for freelancers in 2026.

How do you write a clear AI-use clause?

Keep it short, specific, and boring. A good AI-use clause says what you use AI for, what you do not, who is responsible for the result, and how client data is protected. Boring is the goal here. A clause that reads like a lawyer trying to scare someone makes clients nervous; a clause that reads like a normal operating note makes them relax.

Here is a plain-English version you can adapt:

The freelancer may use AI-assisted tools to support tasks such as research, drafting, and editing. A human reviews and is fully responsible for all final work. The freelancer will not enter the client's confidential information into any public or consumer AI service without written approval. If the client requires that no generative AI be used on any part of this project, the client will state that in writing before work begins, and the freelancer will comply.

Four things make that clause work. It scopes AI to assist tasks, not authorship, so you are not claiming a robot did the thinking. It assigns responsibility to a named human, which is the exact ambiguity the trust research says clients fear. It carves out confidential data, which protects the NDA. And it gives the client a clean opt-out, which respects the buyer who simply wants human-only work.

Match the clause to the regulation when it applies. If your deliverable falls under the EU rules above, add a line that any AI-generated output will be marked as required by the EU AI Act, so the contract and the law agree instead of contradicting each other. For a fuller set of wording options by project type, see our guide to AI clauses in freelance contracts for 2026.

How do you bring it up without scaring the client?

Frame disclosure as a quality and accountability statement, not an apology. The trust penalty in the research comes from ambiguity, the sense that nobody is clearly in charge of the output. Remove the ambiguity and you remove most of the penalty. Say who is responsible, say what stays human, and the AI part stops sounding like a corner you cut.

The phrasing that works is some version of: "I use AI tools to move faster on research and first drafts, and I review and stand behind everything I deliver. Your confidential material never goes into a public tool." That single sentence does the heavy lifting. It names the benefit to the client, which is speed. It names the human accountability, which is you. And it closes the confidentiality worry before the client has to ask.

Notice what it does not do. It does not list every model you touch, and it does not promise the deliverable is 100% machine-free if that is not true. Overclaiming is its own trap. The point is to be accurate and calm, not to perform purity. Experts who study AI disclosure make the same point about customers in general: the disclosure has to be clear and easy to read rather than buried, and it should focus on what actually matters to the person reading it Source.

There is also a reason to expect the ground to keep shifting your way. In that same expert panel, 84% agreed that companies should be required to disclose AI use to customers Source. Norms are moving toward disclosure as the default, not the exception. A freelancer who already discloses cleanly looks ahead of the curve in that world. One who hid it looks like someone who got caught by it.

Where should the clause actually live?

In the contract, not in a chat message or a buried FAQ. A trust-sensitive term that lives in a Slack thread is a term you will struggle to point to when a dispute starts. Put it where both sides signed, and keep the signed copy somewhere you can produce it.

That last part is where a lot of freelancers fall down. The clause is solid, the client agreed, and then six months later the signed version is in an email attachment, the AI-touched deliverable is in a file-share you no longer have access to, and the brief that proves what data you were given is gone. If a confidentiality question ever comes up, you want the agreement and the delivery sitting in one place with timestamps, not scattered across three tools.

Put your AI-use terms in the contract you send, then deliver the work through Delivvo so the signed agreement and the files live in one client portal. If a question ever comes up, both sides see the same record. See how it works →

My bottom line has not moved since the first paragraph. Disclose. Write a short, specific clause. Protect the confidential data above everything. And say it like a professional describing how they work, not like someone confessing. The trust dip from doing this is small and fades as norms catch up. The cost of being the freelancer who hid it, in a year when the EU AI Act is live and client AI policies are everywhere, is not small at all. For the regulatory detail specific to client work, our breakdown of the EU AI Act for freelancers using LLMs on client work in 2026 goes deeper than I can here.