Who Owns the Output If an AI Vendor Trains on My Data?
Dear Will & AiME,
We’re rolling out AI tools across our marketing and product teams, and they’re already using company data to generate content and insights. Our vendor says it may use inputs to improve its models. Should we be concerned about who owns what, or is this just how AI works now?
— Head of Innovation, Chicago
Short Answer 💡
Yes, you should be thoughtful about what you share with AI tools. Many vendors, especially free versions, use your inputs to train their systems, which can affect confidentiality, your ownership rights, and competitive position, even if you technically “own” the output.
Dear Head of Innovation,
AI Vendors Use Your Data Differently Than Traditional Software
Traditional software works like a calculator—you put information in, get a result out, and the software doesn’t remember or learn from what you entered. AI tools are different. They may use what you input and what they generate for you (output) to get smarter over time.
This means your internal memos, marketing plans, product ideas, or customer information could end up embedded in the AI tool.
Some vendors let you turn off this data sharing, especially if you pay for a business or enterprise plan. But the default settings often allow it, and most people never check the fine print. Be sure to check the data privacy settings.
What Rights Do AI Vendors Retain Over Your Data?
The real question isn’t just “Do I own what the AI creates?” It’s “What can the vendor do with the information I shared to get that result?”
Many AI vendor contracts include terms that let them:
Use what you input (and what the AI generates) to make their product better.
Combine your information with data from other users to train their AI.
Keep copies of your questions and the AI’s answers for quality control or safety reviews.
Even if the contract says you “own” what the AI creates, these terms can water down that ownership. The AI might generate something similar for a competitor because your information helped teach it.
It’s also worth asking: Does your confidentiality agreement with the vendor prevent them from using your data to train their AI? And even if they strip out your company name, is there still a risk that your ideas could leak out indirectly?
Output Ownership vs. Input Control: The Real IP Risk
When it comes to intellectual property, there are two things to think about:
Who owns the content the AI creates? This is surprisingly murky. In many places, AI-generated content may not qualify for full copyright protection unless a human was meaningfully involved in creating it.
Who controls the information you put into the AI? You might still technically own your business plans or customer data, but if the vendor uses it to train their system, you’ve lost practical control over it.
This matters especially if your business relies on trade secrets. Once that information is used in a way that could surface elsewhere, even indirectly, it may no longer count as a protected secret.
How to Protect Your Data When Using AI Vendors
Think of AI vendors less like software companies and more like partners who will be handling your sensitive business information.
When reviewing the contract, look for:
Clear language about whether the vendor can use your data to train their AI, how long they keep it, and whether you can opt out.
Whether paying for a business or enterprise plan gives you more protection.
How the vendor defines “anonymized” data, and whether that’s enough protection for your needs.
Inside your organization, consider:
Setting rules about what types of information employees can and can’t enter into AI tools.
Creating simple, clear guidelines so everyone knows what’s acceptable.
Keeping your most sensitive information, like trade secrets or competitive strategies, out of AI tools entirely.
Finally, match your approach to your risk level. For routine tasks (like drafting a generic email or summarizing public information), broader data sharing may be fine. For sensitive work involving your core business strategies, customer data, or proprietary ideas, you need tighter controls.
Before you use an AI tool, ask yourself whether the way the vendor handles your data could undermine the very value you’re trying to create.
— Will & AiME
Three Takeaways:
Many AI vendors can use your data to train their systems. Unless the contract says otherwise, assume what you enter could be shared.
Owning what the AI creates matters less than controlling what happens to the information you put in.
Choosing an AI vendor is a decision about how your data will be handled, not just a software purchase.
