The Question at Hand
A Michigan federal judge recently ruled that a defendant's ChatGPT search history, specifically queries about whether Yahoo could recover deleted emails, demonstrated intent to destroy evidence. The court granted adverse inference sanctions under Rule 37(e), awarded fees, and issued a preliminary injunction based partly on those searches.
This raises a question for every information governance team: When your employees use AI tools to research recordkeeping questions, are you creating a spoliation paper trail?
The answer isn't clear-cut. On one hand, researching data retention capabilities is part of normal IT operations. On the other, courts are starting to view AI query logs as evidence of intent.
The Case for "AI Queries Are Just Research"
Many argue that consulting AI tools shouldn't be treated differently from Googling a technical question or asking a colleague. Under this view, the Conlan Tire case represents judicial overreach.
The defendant asked ChatGPT three straightforward questions: Could Yahoo retrieve an email deleted a year ago under subpoena? How long would a permanently deleted email remain retrievable? Does Yahoo maintain deletion logs? These are factual questions about a third-party system's capabilities. IT staff research these topics routinely when designing retention schedules or responding to legal holds.
If every research query is treated as evidence of bad intent, it penalizes due diligence. Your records manager who Googles "how long does Office 365 retain deleted items" isn't planning spoliation, she's doing her job. The same logic should apply to AI queries.
This interpretation also considers timing ambiguity. The defendant in Conlan Tire queried ChatGPT on June 10, 2025, after litigation began but before the court ordered Yahoo account access on June 24. If he'd researched the same question in May, would it still signal intent? What if he'd asked his IT department instead of ChatGPT? The tool choice shouldn't determine culpability.
Practitioners in this camp recommend continuing normal operations: document your research, maintain clear policies, and don't let fear of discovery chill legitimate technical inquiries. If your retention schedule calls for email deletion after three years, research how your systems implement that schedule. Just document the business reason.
The Case for "Context Matters, and Courts Will Read Intent"
The opposing view holds that Conlan Tire got it right because the queries weren't abstract research, they were targeted, time-sensitive, and followed by immediate deletion.
Judge Kumar noted the defendant asked if Yahoo could retrieve emails deleted "a year ago" (no), then asked how long emails deleted "today" would remain retrievable (seven days), then confirmed Yahoo doesn't log deletions. The sequence matters. He wasn't designing a retention schedule. He was calculating a window.
The forensic timeline reinforced this reading: the court ordered Yahoo authorization on June 24, the defendant provided it by June 27 but delayed account access until July 7, and the forensic exam found no relevant emails. Judge Kumar concluded that if the defendant deleted emails on June 10 and emptied his trash, they'd be unrecoverable by June 17, well before he granted access.
Rule 37(e) requires intent to deprive another party of the information's use in litigation. Intent is inherently contextual. A records manager researching retention in January during policy review looks different from a litigation defendant researching deletion windows three weeks before a court-ordered disclosure.
Courts have always inferred intent from circumstantial evidence: destruction timing, selective deletion, implausible explanations. AI query logs are just another data point. In Conlan Tire, the queries combined with contradictory sworn statements (the defendant claimed he still had his 2022 laptop and phone, but later admitted trading in the phone, and the laptop wasn't manufactured until 2024) painted a pattern.
Practitioners in this camp recommend treating AI queries as discoverable communications. If your team uses ChatGPT or similar tools to research retention, eDiscovery, or legal hold questions during active litigation, assume those queries will be produced. Document the business justification contemporaneously. Better yet, route technical questions through counsel when litigation is reasonably anticipated.
Where Practitioners Actually Land
Most information governance professionals fall somewhere between these poles. They recognize that AI tools are here to stay and that researching system capabilities is legitimate work. But they're also adjusting protocols.
Several organizations now include AI query logs in legal hold scope for custodians under preservation orders. This doesn't mean banning AI use, it means treating those interactions as potentially relevant communications, like Slack messages or emails.
Others are building guardrails: requiring teams to document the business purpose when using AI tools to research retention or deletion during active matters, or routing such questions through legal or compliance channels where the inquiry itself is privileged.
The key shift is treating AI tools as communication channels, not just productivity utilities. When you ask ChatGPT a question, you're creating a record. That record has metadata (timestamp, user, prompt, response). In litigation, it's discoverable.
Our Take
The Conlan Tire ruling doesn't make AI queries inherently suspicious, but it does make them evidence. That's the reality your governance program needs to address.
You can't, and shouldn't, prohibit staff from using AI tools to research technical questions. But you should implement three safeguards:
First, include AI query logs in your legal hold protocols. When you issue a Records Freeze, specify that custodians must preserve AI tool interactions related to the subject matter. This includes ChatGPT histories, Microsoft Copilot logs, and similar platforms.
Second, document business justifications for retention research during active litigation. If your team needs to understand how long SharePoint retains deleted files while responding to discovery, create a ticket or memo explaining why that information is needed for compliance or technical implementation. The research itself is fine, the appearance of furtive inquiry is not.
Third, route legally sensitive questions through counsel. If you're under a preservation order and need to understand deletion mechanics for a specific system, ask your attorney to research it or to document that the research is being conducted at counsel's direction. This doesn't guarantee privilege, but it establishes proper oversight.
The broader lesson from Conlan Tire isn't about AI specifically. It's about intent. Courts have always looked at circumstantial evidence to infer whether a party destroyed records to hide them or pursuant to legitimate policy. AI query logs are now part of that circumstantial picture. Your job is to ensure the picture shows good-faith compliance, not calculated evasion.
If your Records Control Schedule calls for email deletion after seven years, research how Yahoo, Gmail, or Outlook implements that policy. Document it. But if you're researching deletion windows while under a legal hold, expect a judge to ask why.
