Your Product Team May Have Already Given Away the IP

Your product team is using AI to develop your next product.

They are iterating on formulations, refining processes, testing variations. The work is faster than it has ever been. The collaboration feels seamless.

And they may have just given away the IP.

Trade secret protection requires that a company take reasonable measures to keep information secret. That is not a high bar in most contexts. Lock the files. Restrict access. Use NDAs. Courts have generally been practical about what reasonable measures means.

Using a public AI platform to develop proprietary products does not meet that bar.

When your team types a proprietary formulation, a manufacturing process, a product architecture, or a competitive algorithm into ChatGPT, Claude.ai, or any other consumer AI tool, that information leaves your control. It goes to a third party whose terms of service typically permit data collection, storage, and use for model training. The moment it does, you may have destroyed the secrecy that trade secret law requires.

A beauty brand asks public AI to analyze and improve a proprietary formulation. A food company iterates on a recipe. A manufacturer refines a process. A software team debugs proprietary code. Each of these feels like a productivity gain. Each of them may be a trade secret disclosure.

Proskauer Rose published a framework on this in Thomson Reuters in March 2026. Their conclusion was direct. The use of third-party cloud-based AI tools can lead to the inadvertent loss of trade secret protection through user inputs, file uploads, and the potential use of confidential information for model training.

Inadvertent. That is the word that should stop every product leader reading this.

Nobody intended to give away the IP. The team was just trying to move faster.

The patent exposure compounds the problem. Patent protection requires novelty. If proprietary development details entered into a public AI platform influence training data or surface in outputs to other users, the novelty your patent depends on may already be compromised. The filing your legal team is preparing may be protecting something that is no longer protectable.

Last week I wrote about the Heppner ruling and attorney-client privilege. The IP consequence is the same underlying problem with a different face.

Public AI is not confidential. It was not designed to be. The $20 a month subscription your product team is using to accelerate development is a third-party platform governed by terms your legal team has almost certainly not reviewed in the context of what your engineers are actually typing into it.

The distinction between public and private AI is not just a legal and compliance issue. It is a product strategy issue. It is a competitive advantage issue.

It needs to be policy. Not assumption.