Imagine spending years developing a proprietary framework. Something genuinely novel. Something you believe in enough to protect. Then imagine a federal judge looking at your trade secret claim and saying, in plain terms: you gave it away yourself. Not to a competitor. Not through careless conversation at a conference. You gave it away to the AI tool you trusted to help you build it.
That’s not a hypothetical. That’s Trinidad v. OpenAI. And if you’ve ever typed a sensitive idea into a public chatbot, it might already be your story too.
The Case That Should Be Pinned to Every IP Team’s Wall
Sandra Trinidad believed she had something worth protecting. She used the public version of ChatGPT to help develop and refine her proprietary protocols and frameworks — the kind of creative, iterative work that AI tools have made faster and easier for inventors everywhere. Then she tried to assert trade secret protection over what she’d built.
The court dismissed her claims.
Under the Defend Trade Secrets Act, trade secret protection requires that owners take reasonable steps to maintain secrecy. The court found that developing her frameworks through ChatGPT required voluntarily sharing that information with the platform — and that voluntary disclosure was fatal to her claim. It didn’t matter that she intended to keep her ideas confidential. Intention is not a legal defense when you’ve handed your information to a third party that isn’t contractually bound to protect it.
This is what practitioners are now calling the “Trinidad issue.” And it’s landing in IP offices, law firms, and innovation teams like a delayed-detonation device — because millions of people have been doing exactly what Trinidad did, every single day, without once thinking they were creating a disclosure event.
The Uncomfortable Truth About “Private” AI Conversations
Here’s the part that makes IP professionals go quiet when they hear it: consumer AI tools are not confidential by design. They were built for accessibility, speed, and scale — not for the legal standard that trade secret protection requires.
When an inventor opens ChatGPT, Claude, Gemini, or any publicly available AI platform and begins working through an idea, they’re not whispering into a vault. They’re speaking to a third party operating under terms of service that reserve the right to use that data, analyze it, share it with affiliates, and in some cases, train future models on it. The platform isn’t keeping a secret. It was never asked to.
Courts are catching up to this reality fast. Trinidad arrived from the trade secret angle. United States v. Heppner, decided in February 2026 by Judge Rakoff of the Southern District of New York, arrived from the attorney-client privilege angle — ruling that documents a defendant created using a commercial AI tool and then shared with his lawyer were not protected by privilege, because the AI platform’s own privacy policy acknowledged it could disclose user data to third parties. The privilege waiver happened the moment the information touched the platform. Sharing it with a lawyer afterward couldn’t undo it.
Two cases. Two different doctrines. One consistent message from the bench: if the platform isn’t contractually bound to keep your information secret, you’ve disclosed it.
Full stop.
Why This Hits Inventors Harder Than Almost Anyone
The AI industry has built its most valuable intellectual property on top of an interconnected web of data vendors, open-source tools, and shared infrastructure — and that web now constitutes an attack surface that no single company fully controls.
Consider what that actually means. LiteLLM had 97 million monthly downloads. It was trusted because it was useful. The attack didn’t require breaking into Meta or OpenAI directly. It required finding one weak link in a tool used by a vendor those companies relied on. Labs invested billions in compute security and model access controls while treating data vendors as productivity tools. The Mercor incident exposed the cost of that assumption.
EU AI Act requirements around supply chain transparency take effect in phases starting in 2026, with vendor security documentation becoming mandatory for high-risk AI systems. Regulatory scrutiny of third-party AI vendor relationships is arriving whether the industry is ready or not.
The Question Every IP Team Needs to Ask Right Now
Not “do we use AI?” — of course you do, or you will. The question is: what happens to the information after we type it?
Does the platform share user inputs with third parties? Does it train on customer data? Does it operate under contractual confidentiality commitments that would hold up to the legal standard Trinidad and Heppner are now defining? Does your organization actually know the answers to these questions?
Leading practitioners have arrived at a clear position: companies must look to AI tools that maintain the confidentiality of all information inputted and outputted, store data in a private environment unique to the company, and eliminate the concern for shared data or external monitoring by the host. After Trinidad and Heppner, that’s not legal overcaution. It’s the floor.
What Does This Mean for Enterprise IP Teams?
This is the gap that CompassAI, the AI engine inside IQ Ideas+, was designed to close — and it’s worth being specific about why the architecture matters, not just the marketing language around it.
CompassAI does not share customer data with third parties. It does not use customer data to train the underlying model. Data is used only to generate responses and to enhance the user experience within the platform. These aren’t aspirational commitments buried in a terms-of-service footnote — they’re explicit product-level promises, built into the architecture: CompassAI will never share your IP, search topics, or results with another entity, and will never use them for further AI training.
IQ Ideas+ also enforces tenant isolation and controlled internal sharing at the platform level. Ideas are tied to their creators and their organizations. Users can only access what they created or what was deliberately shared within their company. External data transmission follows a minimization approach — only what is essential to generate a response leaves the platform boundary, and nothing more.
Read that back against the Trinidad standard. No voluntary disclosure to a third party. No platform with the right to use, analyze, or share your inputs. Contractual confidentiality that exists before a dispute arises, not as a legal argument constructed after the fact.
That’s not just a better product. Under the legal framework courts are now enforcing, it’s a fundamentally different category of tool.
The Lesson Isn’t “Don’t Use AI”
Sandra Trinidad’s mistake wasn’t that she used AI to develop her idea. Her mistake was using a tool that wasn’t built for what she needed it to do. The same creative, iterative, AI-assisted invention workflow she was running through ChatGPT can happen inside a purpose-built enterprise platform — one where the confidentiality structure exists by design, not by assumption.
Trinidad and Heppner are the opening rulings in what will be a long judicial conversation about AI, IP, and disclosure. More cases are coming. The legal direction is clear. And the gap between a consumer chatbot and a purpose-built IP platform is no longer just a product differentiator.
It’s the difference between owning your invention and accidentally giving it away.
Trinidad and Heppner make one thing clear: the legal risk of using the wrong AI tool in your invention workflow is no longer theoretical. IQ Ideas+ with CompassAI was built for teams who can’t afford to treat that risk as an acceptable unknown — with contractually enforced confidentiality, no third-party data sharing, and enterprise access controls that reflect what courts are now saying the standard should be. Request a demo to see IQ Ideas+ in action.
