How Should I Structure Vendor Contracts for AI Use?

Dear Will & AiME

We're starting to bring in vendors that use AI tools to deliver services. How can we structure our contracts to enable the use of AI but still protect our IP, data, and business?

— Alex in Illinois

Short answer💡

Structure AI vendor contracts to clearly define permitted AI use, secure full rights to outputs, and restrict how your data is used or shared with AI tools. The key is controlling inputs, outputs, and vendor behavior to prevent IP loss, data exposure, or unexpected licensing issues.

Dear Alex in Illinois,

Working with vendors who use AI tools to support your business can offer speed and scale, but it also introduces a very real risk: what happens when their AI usage crosses lines you didn't intend?

Whether you're concerned with ownership of deliverables, rights to training data, or trade secret exposure, a vendor agreement is your strongest shield. But only if you build it right. Let's walk through the key pieces every AI-conscious vendor agreement should cover.

Four Clauses Every AI-Related Vendor Contract Should Have

1. Clear Authorization of AI Tool Use

Don't leave authorization open-ended. The contract should identify:

• Whether the vendor is allowed to use AI tools at all,

• Which specific tools are permitted (e.g., named vendors or approved categories),

• Whether those tools can be used for all tasks or only specific parts of the engagement.

This gives you control over how AI enters your business, rather than finding out after the fact that your confidential white paper was drafted using an unvetted chatbot.

2. Commercial Rights to AI-Generated Output

If the vendor is using AI to help generate your deliverables, make sure:

• You're getting full commercial rights to use, modify, distribute, and monetize the output.

• There's no upstream restriction from the AI tool (e.g., terms of service that prohibit commercial use or require attribution).

The contract should confirm that the vendor has cleared all rights and is granting you the output free of encumbrances.

3. No Assignment of Your Rights Back to the AI Provider

Many AI tools bake into their terms that user prompts and data can be reused to improve the model. You want to ensure the vendor:

• Is not submitting your confidential materials, branding, code, or strategic documents into an AI tool that claims any license or rights over the input.

• Agrees not to use tools with "improvement" or "reuse" clauses without your written approval.

This is especially important if the deliverables relate to anything proprietary or regulated.

4. Ban on Submitting Trade Secrets to AI Tools

Your vendor should contractually agree not to input your confidential or trade secret information into any AI system—period. Even "private" versions of AI tools may not be truly secure. This clause should:

• Define what constitutes sensitive or confidential inputs,

• Prohibit such inputs into public or third-party AI systems,

• Impose liability if the vendor violates this rule.

You Can Enable AI Use Without Losing Control

To be clear: you can enable vendors to use AI while protecting your business. But the contract has to do more than say "vendor will follow applicable law." It should establish real guardrails, reflect how AI tools work today, and anticipate potential areas of concern.

AI risk is a complex area that concerns IP ownership, data leakage, licensing rights, and brand control. Tight contracts give you leverage, clarity, and peace of mind.

-Will & AiME

Three Takeaways:

1. Require vendors to get approval for AI use and identify which tools are allowed.

2. Ensure you receive full commercial rights to AI-generated deliverables and that the tools used don't claim ownership.

3. Prohibit inputting trade secrets or confidential data into AI tools without express written consent