Your design team just used AI to draft the concept documents. Your project manager ran the RFP response through ChatGPT. Your consultant generated three specification sections in an afternoon.

Who owns any of that?

This is not a hypothetical. It is a question your legal team should already be asking, and if they are not, you are exposed.

The law finally caught up (sort of)

In January 2025, the U.S. Copyright Office clarified where AI fits. The short version: AI is treated like any other software tool. Use Microsoft Word to write a contract, you own it. Use Revit to design a building, you own it. Use AI with meaningful human input and editing, you own it.

The key phrase is “meaningful human input.”

Enter a single prompt into ChatGPT and publish whatever comes out? Do not expect a copyright to protect you. But if someone crafts the prompt carefully, edits the output, builds on the firm’s existing work, and applies professional judgment throughout, that is a different story.

The part that actually matters to you

What building owners consistently miss in this conversation is that they sit on both sides of the issue.

You hire design and construction firms. They are now using AI to produce deliverables faster. That is mostly good news for your schedule and budget. But it raises a question worth pressing in every contract negotiation: what are the IP implications of AI-generated work product?

If your architect used a generic AI prompt to produce a specification and another client received the same output, who owns what? If a contractor ran your project data through an AI tool with permissive terms-of-service, where did that data go?

You have a right to understand how your vendors are using AI and what that means for the work product they deliver to you.

The go-by principle actually works here

The AEC industry has always run on “go-bys”: reusing past proposals, specs, and designs as starting points. It looks like laziness from the outside. It is actually sound risk management.

In the AI era, that habit pays off in a new way. The most defensible AI-generated work is work that begins from something you already own: your previous project documents, your firm’s past RFP responses, your existing specifications.

AI augmenting your existing work sits on solid ground. AI generating from scratch with no human judgment layered in sits in gray territory, legally and professionally.

When evaluating vendors, this distinction matters. Firms that use AI thoughtfully (starting from their own IP, adding genuine human expertise, and documenting the process) are building on a defensible foundation. Firms that allow AI to write deliverables from scratch without meaningful professional review are accumulating liability.

What this means for your projects

Ask your design and construction partners three things:

Do you have an AI policy? Any firm doing serious work in 2025 should have one. If they look at you blankly, that is informative.

Where is my project data going? When your project documents get fed into an AI tool, what are the terms? Is that data being used to train future models? Who has access?

How much human review happens before I see a deliverable? There is a meaningful difference between AI-assisted work and AI-generated work. The first has a professional standing behind it. The second may not.

The shift that is coming

AI is being embedded directly into the tools your vendors already use: Revit, AutoCAD, BIM platforms. The line between “human work” and “AI work” will blur further every year.

That does not warrant alarm. It warrants staying informed and asking more precise questions. Copyright law is catching up to practice. Your contracts should do the same.

The best outcome is not zero AI. It is AI used with intention, where the human expertise your vendors are paid for actually shows up in the work. Technology that makes your project better, faster, and more defensible.

That is the standard. Make sure your vendors understand you are holding them to it.