Conlan Tire Co. v. Gonzales: When a Litigant’s ChatGPT History Becomes Evidence of Intent

Collect your client's AI chats before the other side does: a Michigan court used them to prove intent

In October 2026, ACEDS republished a Minerva26 Blog analysis of Conlan Tire Co. v. Gonzales, a federal case in Michigan. Conlan hired Alec Gonzales as an account executive in July 2022 under a confidentiality and non-solicitation agreement. It fired him in December 2024 and sued in March 2025, after he joined a direct competitor. His only defence was that he never signed the agreement. A stipulated ESI order in June 2025 gave Conlan's forensic examiner access to his Yahoo account and his 2022 devices. The examination of his devices found ChatGPT questions, asked earlier that month, about whether Yahoo could recover a deleted email even under subpoena. The court found he intended to deprive Conlan of evidence and imposed an adverse inference, which helped support a preliminary injunction against him.

Does a litigant's AI chat history count as discoverable evidence?

Yes. It is stored on phones, laptops and accounts, timestamped and written in the user's own words, so it is ordinary ESI. Here the court used it to infer state of mind.

The court read what Gonzales asked, not what ChatGPT answered. The questions were whether Yahoo could retrieve a year-old deleted email, how long a permanently deleted one stays recoverable, and whether Yahoo logs it. The repost doesn't say whether the answers were accurate, and the intent finding doesn't need them to be. For discovery purposes, the prompt is a statement of what the user was thinking.

The Minerva26 post puts it bluntly: AI chat history "is evidence of state of mind." Earlier rulings in its series asked whether a party's AI use is protected. This one asks what the use proves.

What should change in preservation and collection protocols?

AI chat history should appear in preservation letters and forensic collection scopes, and clients should be told how to behave before they use AI on their matter.

That second step is the cheap one. The chats in Conlan were created after the suit was filed, in the middle of discovery. The post notes that clients now ask AI before they call counsel, after they hang up, and during discovery. Any protocol that stops at email, chat and text will miss a source the other side may now request.

How much weight should one ruling carry?

Limited weight. It is a single federal ruling at the injunction stage, reported through a practitioner blog repost, and the surrounding conduct was unusually poor.

Gonzales gave Yahoo access late. He produced a laptop but not his 2022 iPhone, which he now said he no longer owned. His second declaration had said he still had both 2022 devices. The chats were damaging in that context. The ACEDS text I read breaks off mid-account, so I can't say how the court weighed the chats against the rest. Treat it as a warning about one more data source, not a settled standard for AI evidence.

What should I ask my e-discovery vendor?

Ask whether the platform can collect, preserve and review AI chat histories from client devices and accounts, with timestamps and chain of custody intact, and where that sensitive material is processed.

Specifically: which AI apps and export formats are supported, and are conversations kept as threads or flattened into loose messages? Can a reviewer cite a prompt by its timestamp? A client's AI chats may show case strategy and fears, so ask whether that data touches any model training. Also ask whether you can review your own client's chats early. Opposing counsel may ask for them.

Frequently asked questions

Did the court rely on ChatGPT's answers?

The sources describe reliance on the questions, as evidence of intent. They don't say whether the answers were correct.

Was the adverse inference the whole basis for the injunction?

No. The source says it helped carry the injunction. The text available doesn't set out the other grounds.

Should clients stop using AI on their cases?

The post doesn't call for a ban. Its advice is to talk to clients first, so they know the chats are discoverable.

Sources: ACEDS, republishing the Minerva26 Blog.