Can I legally prevent clients from using my work to train AI?

Yes, you can prevent this by explicitly prohibiting the use of your deliverables for machine learning or AI model training in your Master Services Agreement (MSA) or Statement of Work (SOW).

What specific language should I add to my contracts?

You should include a 'Prohibition on AI Training' clause that restricts the client from using, processing, or inputting your work into any generative AI model or machine learning dataset.

A TermScore is an AI-generated risk rating from 0 to 100 that evaluates the fairness and risk level of any legal document. A score of 86-100 means the contract is highly favorable, while 0-30 indicates dangerous terms you should walk away from.

What types of documents can TermScore analyze?

TermScore analyzes any legal document including freelance contracts, employment agreements, residential leases, NDAs, brand sponsorship deals, SaaS vendor agreements, and rental applications. Upload PDFs, Word documents, or images.

How much does TermScore cost?

TermScore offers a free plan with 3 scores per month. The Team plan is $15/month for 30 scores, and the Professional plan is $50/month for 500 scores with team collaboration features.

Is my document data secure?

Yes. Documents are processed in memory and analysis results are stored encrypted. TermScore never shares your document data with third parties and you can delete your data at any time from your dashboard.

How to legally protect freelance source files from unauthorized AI model training by agency clients

Protecting Your Freelance Assets from AI Training

To legally protect your freelance source files from unauthorized AI model training, you must include an explicit 'No-AI Training' clause in your contracts. Standard 'Work for Hire' language is no longer sufficient; you must restrict the client’s right to use your deliverables as input data for machine learning systems.

The Legal Reality of AI Data Scraping

Most standard freelance contracts grant the client 'all rights, title, and interest' in the work product. In the current legal landscape, this broad assignment of rights is being interpreted by some agencies as a license to feed your work into internal or third-party AI models. Without specific limitations, you are essentially providing the raw material for your own professional replacement.

Why Standard Clauses Fail

Work for Hire: Transfers ownership but does not explicitly forbid the *method* of processing the data.
Perpetual License: Often grants the client the right to use the work in 'any media now known or hereafter developed,' which includes AI training sets.
Confidentiality Gaps: Many NDAs do not define 'inputting data into a Large Language Model' as a breach of confidentiality.

Key takeaway: If your contract does not explicitly mention 'machine learning' or 'AI training,' assume the client believes they have the right to use your files for those purposes.

Action Item: Review your current MSA for any mention of 'derivative works' or 'data processing' rights. If these terms are present without an AI exclusion, you are at risk.

Drafting the 'No-AI' Clause

Your contract must define the scope of usage. You do not necessarily need to withhold ownership, but you must withhold the right to use the work for model training.

Essential Components of an AI-Protection Clause

Definition of Prohibited Use: Explicitly name 'machine learning,' 'generative AI,' 'neural network training,' and 'automated data processing.'
Scope of Restriction: Apply the restriction to both the final deliverable and the underlying source files (e.g., PSDs, code repositories, raw footage).
Indemnification: Require the client to indemnify you if they breach this term and your work is subsequently leaked or used in a public model.

Clause Type	Effect on AI Training	Risk Level
Standard Work-for-Hire	Allows AI training	High
Limited License	Restricts use to specific project	Medium
Explicit AI-Prohibition	Legally bars AI training	Low

Action Item: Draft a rider titled 'AI Usage Restrictions' and append it to your existing SOWs. Ensure it states that any breach of this clause constitutes a material breach of the agreement.

Jurisdictional Considerations

While contract law is generally enforceable, the legal status of AI training varies. In the United States, the 'fair use' doctrine is currently being litigated regarding AI training. However, contract law supersedes fair use; if a client signs a contract agreeing not to use your work for AI, they are bound by that contract regardless of whether the law would otherwise permit them to scrape your work.

Checklist for Contract Audits

Does the contract grant the client the right to 'process' the data?
Is there a 'Grant of Rights' section that includes 'all purposes'?
Are there specific carve-outs for 'internal research and development'?

Key takeaway: A contract is your strongest defense. Even if copyright law is ambiguous regarding AI, a signed contract prohibiting AI use creates a clear path for litigation if the client violates your terms.

Action Item: If a client refuses to sign an AI-prohibition clause, consider increasing your project fee by 20-30% to account for the loss of exclusivity and the potential devaluation of your work through AI automation.

Using Technology to Enforce Your Rights

Manually reviewing every contract for hidden AI-training loopholes is time-consuming and prone to human error. TermScore automates this process by scanning your agreements for problematic language, identifying 'AI-training' risks, and suggesting specific, enforceable clauses to protect your intellectual property. By using TermScore, you ensure that your freelance contracts are as robust as your work product, allowing you to focus on creation rather than legal defense.