Do I Own the Source Code of My AI Bot?

Your AI bot is live, users love it, and then a lawyer asks: "Who actually owns this code?" The answer depends entirely on who built it and under what contract—and getting it wrong can cost you the asset entirely.

This is one of the most underasked questions in AI software procurement. Whether you hired a freelancer, a dev shop, or used a no-code platform, the default answer is almost never "yes, you own it automatically." Here's what you need to know before you sign anything—or after you already did.

What "Owning Source Code" Actually Means

Intellectual property (IP) ownership in software is governed by copyright law. The moment code is written, copyright vests in the author—not the buyer, unless a specific legal instrument transfers it.

There are two main ways ownership reaches you as a client:

• Work-for-hire agreement: If the developer is your employee or a contractor who signs a written work-for-hire clause, the copyright transfers to you automatically.
• IP assignment clause: A contract provision that explicitly assigns all rights—source code, models, prompts, documentation—from the vendor to you upon delivery or payment.

Without one of these in place, you likely have a license, not ownership. That's a critical distinction.

License vs. Ownership: The Practical Difference

If your AI bot runs a core business workflow—customer support, lead qualification, internal knowledge retrieval—operating on a license is a structural business risk.

The Three Scenarios Where You Probably Don't Own Your AI Bot's Code

1. You Built It on a SaaS No-Code Platform

Tools like Intercom, Drift, Chatbase, or ManyChat let you configure an AI bot through a visual interface. You own your data and conversation flows, but the underlying code, models, and infrastructure belong entirely to the platform.

If that platform changes pricing, kills the product, or gets acquired, you have zero recourse. You cannot export the bot and run it elsewhere.

2. You Hired a Dev Shop Without an IP Assignment Clause

Many software agencies—especially those using offshore teams or subcontractors—retain IP by default unless their contract explicitly says otherwise. A deliverable spec is not the same as an IP assignment.

Red flags in contracts:

• Language like "license to use the deliverables" instead of "assigns all rights"
• No mention of open-source components and their licenses
• Vague ownership of "custom work" without defining what that includes

3. You Used a White-Label or Embedded AI Product

Some vendors sell AI bots as a managed service where they host, maintain, and update the model. You get a product interface, not a codebase. The moment you stop paying, the bot disappears.

This model can work for simple use cases, but it's not a software asset—it's a subscription to someone else's asset.

Do I Own the Source Code of My AI Bot If I Used OpenAI or Another LLM API?

This is where it gets nuanced. If a developer built a custom application that calls the OpenAI API (or Anthropic, Google Gemini, etc.), the application layer—the code that routes requests, manages context, stores embeddings, handles authentication—can absolutely be owned by you.

What you don't own:

• The underlying foundation model (GPT-4, Claude, etc.)
• The training data or weights
• The API infrastructure

What you can own:

• The custom orchestration logic (LangChain flows, agent loops, tool calls)
• Your prompt engineering and system prompts
• The retrieval-augmented generation (RAG) pipeline and vector database setup
• Fine-tuning datasets you created
• The application UI and backend

This is exactly why how the contract is written matters as much as how the software is built.

How to Verify What You Own Right Now

If you already have an AI bot in production, run through this checklist:

Review your contract:

• Does it contain an explicit IP assignment clause?
• Is there a work-for-hire provision?
• What does it say about open-source third-party components?

Request a code audit:

• Ask your vendor for access to the full repository (GitHub, GitLab, etc.)
• Verify you have admin-level access, not just read access
• Check that the repo is not forked from a licensed or restricted base

Check your hosting and infrastructure:

• Is the bot deployed on your cloud account (AWS, GCP, Azure) or the vendor's?
• Who holds the API keys for the LLM provider?
• Who controls the domain and SSL?

Assess dependency risk:

• Which open-source libraries are used? MIT and Apache 2.0 are permissive; GPL may impose obligations on derivative work.
• Are there any proprietary SDKs embedded in the codebase?

What Full Code Ownership Should Look Like in a Contract

When negotiating with an AI development partner, push for these specific provisions:

1. Full IP Assignment: "Upon final payment, all intellectual property rights in the deliverables, including source code, documentation, and training data, are irrevocably assigned to the Client."
2. No Retained License: The vendor should not retain a license to reuse your code in other projects.
3. Third-Party Component Disclosure: A schedule listing all open-source libraries and their licenses.
4. Repository Transfer: The vendor transfers the Git repository—including full commit history—to a client-controlled account at project close.
5. No Recurring Fees for the Software Itself: Distinguish between infrastructure costs (which you'd pay to AWS or OpenAI directly) and software licensing fees (which should be zero).

Why This Matters More for AI Bots Than for Traditional Software

AI bots are not static applications. They involve:

• Prompt engineering that directly shapes the product's behavior
• Fine-tuned models or embeddings trained on your proprietary data
• RAG pipelines that give the bot access to your internal knowledge
• Feedback loops that improve the model over time using your users' interactions

Each of these layers represents accumulated business value. If you don't own them, you're building equity on someone else's land.

A competitor who builds an equivalent bot with full IP ownership can iterate, sell, license, or pivot the product. One who operates on a vendor license cannot.

The Catalizadora Approach: 100% Ownership, No Recurring Licenses

At Catalizadora, every engagement is structured around a simple principle: you own everything we build. Full IP assignment is standard in every contract—not an add-on, not a negotiation point.

When we deliver a custom AI application—whether through a 12-week full build (Core), a 15-day focused sprint (Solo), or a scoped custom engagement (Forge)—the client walks away with:

• Full source code in a client-controlled repository
• Complete transfer of all custom models, prompts, and pipelines
• Zero recurring license fees to Catalizadora
• Documentation sufficient to hand off to any development team

You pay for the build. You own the asset. Ongoing costs are only the infrastructure you choose to run it on—your cloud, your LLM API keys.

This is what software ownership should look like in 2025.

CTA: Build AI You Actually Own

If you're evaluating an AI bot project—or questioning what you actually own from a past build—start with clarity on IP.

Read the Catalizadora Manifesto →

It outlines exactly how we think about software ownership, client relationships, and what "building for your business" really means. No recurring licenses. No vendor lock-in. Just software that belongs to you.

Summary: Key Takeaways

• Copyright in code vests in the author by default—clients need an explicit IP assignment to truly own the software.
• No-code platforms, white-label bots, and poorly structured agency contracts almost never transfer full ownership.
• You can own the application layer of an AI bot even when it uses third-party LLM APIs.
• Full ownership means: client-controlled repository, no retained vendor license, no recurring software fees.
• Before signing any AI development contract, verify the IP assignment clause, open-source disclosures, and repository transfer terms.