Most litigators believe they are skilled investigators, and many are right. Depositions, document requests, and cross-examination are investigative tools, and good trial lawyers use them well. But the confidence that comes from being good inside a case has a blind spot: it obscures risks that arise the moment an attorney, or the attorney’s staff, starts doing the kind of background digging that often belongs to a third party. None of these risks require sloppiness. They show up even when the work is done carefully.
The first is a discovery problem, not an investigation problem. When an attorney or a paralegal under the attorney’s direction gathers facts directly — calling a former business partner, pulling corporate filings, compiling a witness’s litigation history — that work is typically protected, if at all, only as fact work product under Federal Rule of Civil Procedure 26(b)(3), a qualified protection that opposing counsel can overcome by showing substantial need and undue hardship.
Courts have also found that a paralegal’s fact-finding, even when supervised by counsel, falls outside the attorney-client privilege entirely, since privilege attaches to communications, not investigative legwork (Segway, Inc. v. Special Olympics, Conn., 2015 Conn. Super. LEXIS 2711). The underlying facts an attorney’s own office gathers are more exposed than many assume when they decide to “have someone look into it” rather than engage an outside investigator whose work sits at arm’s length from the litigation team’s own files.
The second risk is an ethics trap sitting in territory attorneys consider routine: a witness’s or party’s online footprint. Reviewing what someone has made genuinely public is uncontroversial; the trouble starts at the boundary between public and private, which is exactly where the useful information tends to live. ABA Model Rule 4.1(a) bars knowingly making a false statement of material fact to a third person; Rule 8.4(c) bars dishonesty, fraud, or misrepresentation generally; and Rule 5.3(b) extends both to anyone acting in the lawyer’s direction, including investigators.
The Philadelphia Bar (Opinion 2009-02) and the New York City Bar (Formal Opinion 2010-2) have both held that a “friend request” made without disclosing its connection to litigation is deceptive conduct under these rules, regardless of how relevant the resulting information is. An Ohio prosecutor was suspended for twelve months after creating a fake profile to elicit statements from a defendant and a witness. The rules governing deceptive access apply equally to the attorney and to anyone the attorney directs.
The third risk surfaces later, when there is no time left to fix it: authentication and foundation. Competent representation under Rule 1.1 includes considering, at the time information is gathered, how it will be authenticated and admitted later. Material an attorney’s own staff collects informally — a screenshot, a printout, a summary memo — routinely runs into foundation objections that material from an independent, documented source does not.
A fourth risk has emerged more recently, and it compounds the others rather than sitting apart from them: a growing number of attorneys are turning to generative AI to do the same background work, and the temptation is obviously fast, cheap, and the output reads like finished work product. But speed is not free. On accuracy, a federal expert witness’s report drafted in part with ChatGPT was found to contain fabricated citations and invented testimony serious enough that the court said the errors “shattered his credibility,” and the case was dismissed (U.S. ex rel. Khoury v. Intermountain Healthcare, D. Utah, dismissed Sept. 30, 2025).
On privilege, the exposure may be worse. In United States v. Heppner, 2026 WL 436479 (S.D.N.Y. Feb. 17, 2026), the court held that documents a defendant generated using a public AI platform, then shared with his attorneys, were protected by neither privilege nor work product — because inputting information into a consumer AI tool whose terms permit third-party data sharing destroys the confidentiality privilege depends on, and because work the attorney did not direct cannot be the attorney’s work product.
A tool that cannot be supervised the way a paralegal can, and cannot guarantee where its inputs end up, is a different category of risk than a directed human investigator whose engagement and confidentiality terms are defined before the work begins.
None of these argues that attorneys investigate badly. It argues that investigative work at an attorney’s own office generates different discovery exposure, ethical exposure, and evidentiary weight than the same facts gathered by an independent third party — a difference that becomes apparent only when opposing counsel first finds it. The practical takeaway is not less investigation, but deliberateness about which parts belong in-house, and which are better handled at arm’s length, documented, and insulated from the litigation team’s own exposure from the start.
