Human Resources and other business leaders, using or allowing workforce members to CHAT-GPT or other artificial intelligence (“AI”) tools to research, make decisions or to support other activities should ensure that their organizations and their teams understand and manage the resulting evidentiary consequences and responsibilities these activities create.
Why AI Searches May Be Evidence
AI prompts, outputs, and related metadata often qualify as discoverable electronically stored information (“ESI”) in litigation, regulatory audits, and enforcement proceedings.
Under the Federal Rules of Civil Procedure, discoverable information includes electronically stored information (ESI”) relevant to claims or defenses. See Fed. R. Civ. P. 26(b)(1); 34(a)(1)(A). Once litigation is reasonably anticipated, organizations must preserve relevant ESI under Federal Rules of Civil Procedure. See e.g.,Hoffer v. Tellone, 128 F.4th 433 (2d Cir. 2025).
AI searches and other interactions and the information and other outputs they produce generally qualify as ESI for purposes of the Federal Rules of Civil Procedure, Federal Rules of Evidence, (hereafter collectively the “Federal Rules”) and comparable federal and state litigation procedural rules.
Likewise, federal and state regulatory and enforcement agencies tend to consider AI and other ESI evidence covered by document retention and discovery rules.
Where applicable, these Federal Rules and agency rules generally include AI and other ESI evidence organizations must preserve, identify, and subject to discovery or other production like traditional evidence in litigation and agency audits and investigations.
ESI evidence generally includes any data stored in electronic form—such as emails, texts, spreadsheets, social media, and Internet of Things (“IoT”) data. As broadly construed by the courts, courts already long have admitted:
- Internet search histories;
- Internal chats and Slack messages;
- Draft documents; and
- Deleted files.
AI searches are simply the next evolution of this evidence category. AI records and information considered ESI can include;
- AI prompt histories;
- Generated outputs;
- Embedded AI-assisted drafts;
- Platform logs (if accessible); and
- Other records.
Organizations that fail to fulfill requirements for AI or other ESI early identification, data authentication and other requirements of Federal Rule of Evidence 902, the requirements of Federal Rule of Civil Procedure 37(e) regarding lost evidence, or other applicable requirements to preserve and produce ESI may result in sanctions, adverse inferences, and penalties under the Federal Rules. Similarly, organizations may incur evidentiary sanctions, regulator penalties, and other adverse consequences for failing to identify, retain and produce ESI from AI or other sources in government audits and investigations. Common sanctions include:
- Monetary sanctions;
- Evidence preclusion;
- Adverse inference jury instructions; and
- Other authorized sanctions.
See, e.g., Jones v. Riot Hosp. Grp. LLC, 95 F.4th 730 (9th Cir. 2024) (affirming dismissal as sanction for intentional destruction of ESI); Maziar v. City of Atlanta, No. 1:21-cv-02172, 2024 WL 197561 (N.D. Ga. June 10, 2024) (denying summary judgment and awarding fees based on loss of text messages); McBride v. Moore, No. 2:23-cv-02904, 2024 WL 1136429 (C.D. Cal. Feb. 23, 2024) (denying sanctions where ESI not shown lost or duty not triggered). Gregory v. State of Montana, No. 22-____ (9th Cir. 2024) (reversing sanctions imposed outside Rule 37(e); emphasizing Rule 37(e) as exclusive remedy for ESI spoliation); DR Distribs., LLC v. 21 Century Smoking, Inc., 513 F. Supp. 3d 839 (N.D. Ill. 2021) (recognizing financial prejudice from spoliation and awarding fees); Bistrian v. Levi, 448 F. Supp. 3d 454 (E.D. Pa. 2020) (Rule 37(e) provides exclusive framework for ESI spoliation); Fast v. GoDaddy.com LLC, 340 F.R.D. 326 (D. Ariz. 2022) (failure to preserve mobile device data warranted sanctions).
These and other related cases alert organizations that AI and other modern data sources are squarely within ESI. Courts treat AI, texts, mobile data, and app-based communications as discoverable ESI.
The precedent reflects that the best opportunity to position your organization to show the reasonability of the actions taken is through the existence and enforcement of policies before and during the use of the AI tool.ESI preservation obligations depend on foreseeability, control, and access to data. When deciding the consequences of the unavailability or failure to produce ESI, the determination regarding failure to take “reasonable steps” is fact-intensive. Of course, the failure to retain the documentation will be particularly likely to be found unreasonable where the party was under a statutory, regulatory, ethical, contractual, or other pre-existing obligation to preserve the evidence.
Also, courts require proof of intent to deprive before imposing the most severe adverse inference or dismissal sanctions, a lack of proof of intent to deprive the request requesting party of evidence does not mean there will not be consequences for the non-producing party. Negligent failure to retain and produce ESI and other evidence still carries consequences. Even without bad faith, courts may impose curative measures, fees, or evidentiary limitations.
Similarly, the U.S. Department of Justice Securities and Exchange Commission Department of Health and Human Services Office for Civil Rights Equal Employment Opportunity Commission and other federal and state government agencies are increasingly sophisticated in digital evidence collection.
Among other things, organizations should be prepared to routine and produce documentation and data obtained, utilized, or otherwise interacting with AI tools and it’s associated meta-data and other components to respond to litigation and regulatory request for a broad range of data and information. Optimally the data captured and retained should include, but it’s not necessarily limited to:
- AI usage policies;
- Employee and other agent AI interaction records;
- Evidence of AI and other relevant governance and training; and
- Data protection controls.
Consequently, failure to govern, identify, preserve and produce AI-generated and AI-assisted records appropriately can expose organizations to spoliation sanctions, adverse inference instructions, regulatory penalties, loss of privilege protections, and expanded liability exposure.
In recognition of the possibility that AI tool interactions may give rise to obligations to retain and produce ESI evidence created as a consequence of that interaction, or organization should work with legal counsel to develop an administer appropriate practices to monitor, identify, retain, manage, and where necessary produce this ESI evidence.
AI Tools Create Evidence
Beyond considering and meeting documentation and other evidently protection, preservation, and production responsibilities, corporations also should give some thought to the very nature of the evidence created through the use of AI tools.
Organizations should keep in mind that the use of AI tools creates legal evidently risks because AI tools typically generate synthesized responses (not just links) that often incorporate user inputs into outputs that may reflect user intent, knowledge, biases and opinions, and decision-making. This often makes AI interactions particularly valuable evidence for:
- Intent (e.g., “how to terminate employee without legal risk”) Knowledge (awareness of compliance obligations) State of mind (deliberate vs. negligent conduct);
- Knowledge (awareness of compliance obligations);
- State of mind (intentional, willful, or deliberate vs. negligent conduct)
The potential risks of this and other evidence is heightened by the fact that the evidence created may arise not only from the actions taken by the user of the AI tool, but also may be inherently built into the design of the AI tool itself or the databases or other reference materials that it accesses, not all of which may be transparent to the user or the organization that employs the user. These risks are further heightened when the AI tool use is not conducted internally within the organization by its employee, but rather is a tool utilized by a consultant or other third-party provider conducting activities of a sensitive nature on behalf of the organization, such as a recruiting company, investigation, company, or other service provider.
These unique characteristics of AI make it advisable that organizations recognize and manage potentially heightened exposures that employee or other agent use AI tools can produce for the organization in a wide range of sensitive areas.
Examples of queries that can become “smoking gun” evidence include but are not limited to:
- In employment or other workforce administration searches, AI queries such as “How to terminate employee with medical condition,” ‘How to avoid claims when, terminating older, disabled, complaining, injured or other employee with protected status, or the like can be evidence of discriminatory or other adverse intent;” or “How to beat a union organizing campaign;”
- Compliance & Regulatory searches such as “How to structure payments to avoid reporting requirements,” “HIPAA penalties for disclosure; searches about compliance or looking for compliance loophole; searches where company researched sanctions for noncompliance in areas involved in litigation or enforcement; or searches on risk management that could be evidence the organization saw but chose not to follow rules or standards or otherwise looked for or acted to circumvent compliance or disregarded interpretations less favorable to chosen challenged course of action;
- Litigation or Other Defense Strategy searches or tools such as “How to defeat a whistleblower claim,” “Ways to minimize damages in lawsuit” “Protecting your assets from IRS or in bankruptcy,” “How to conceal” or How to hide” orthe like can harm the organization’s interest by showing adverse intent, willfulness, or other motive or state of mind;
- Litigation case law, enforcement, argument drafting, or other actions that could reveal or provide insight on sensitive litigation strategies or their strengths or weaknesses;
- Financial & Tax searches such as “Aggressive tax strategies unlikely to be audited,” or “contract terms to reclassify employee to contractor,” “Structure transaction to avoid disclosure” or the like; and
- Other searches or tool uses that could reflect improper, intent, or document improper activities, such as how to hide evidence, how to create a bomb, how to poison somebody or that creates a record of conduct such as edits to revise data or documentation in reports or records, where the changes are tracked and retained.
Given these other risks, organizations should carefully consider and manage these and other risks when deciding whether, when, how and what AI tools their organizations allow their people to use, who gets to use what tools, designate and train those authorized to use these tools appropriately, and design and implement appropriate tools to track, capture, retain and manage these records of AI use and their implications. Optimally, the planning should identify and work to manage the creation and preservation of evidence and related AI ESI required or otherwise helpful to meet, applicable, regulatory, contractual, statutory, or other requirements in a manner that minimizes the creation of evidence that could call into question the compliance or other appropriateness of the organizations actors.
Privilege and Confidentiality Risks
Using AI tools and output also can have implications on the ability of an organization to protect legal advice and work product developed and shared within the scope of attorney privileges from discovery in judicial or regulatory actions. Organizations need to recognize risk to the confidentiality of legal advice or work product that entering sensitive legal questions or information into public AI tools not specifically designed and used outside the scope of the attorney-client relationship to avoid creating problematic evidence, disclosing discussions or work product that otherwise might qualify for protect against discovery in litigation or agency proceedings under the attorney-client privilege or attorney work product rules, or both.
Searches conducted by organization employees, consultants, or other agents or representatives about the law, strategies, or legal risks and consequences without or outside the scope of an attorney-client relationship generally can be discovered and used as evidence. Consequently, organizations should regulate the use by officers, directors, compliance officers, human resources directors, consultants, non-legal investigators and auditors and others of AI tools, internet or other searches to investigate the law or legal strategies independent of or outside the scope of attorney-client privilege.
Particularly risky scenarios include:
- In-house counsel, Human Resources, risk management or compliance staff using public AI tools;
- Employees seeking legal guidance outside approved channels; or
- Uploading contracts, PHI, or proprietary data into AI systems.
Additionally, organizations and others communicating or working with legal counsel on behalf of the organization within the scope of attorney-client privilege to design strategies or investigate or defend actions generally should not use AI tools to conduct their own legal research or analysis without authorization and direction of the legal counsel to avoid forfeiting attorney-client privilege and work product protections.
If the required confidentiality is preserved, the attorney-client privilege and work product privileges rules can protect confidential communications between a client and its attorney and work product prepared for risk management, defense or other purposes of the legal engagement against disclosure in litigation or other proceedings in many circumstances. However these protections are lost if the communication or work product is disclosed to or discussed with third parties outside the attorney-client relationship. Entering factual information, conducting legal searches, or using AI tools outside the attorney-client relationship, not specifically designed to preserve confidentiality, or both to draft or evaluate legal documents, research, drafts, or strategies generally is considered a third party disclosure that can waiver or undermine the privilege for the specific information input to the AI tool as well as potentially related communications or work product.
For these and other reasons, organizations and individuals generally should resist the temptation to use AI tools to evaluate legal strategies, advice or work product.
HIPAA, Trade Secret, and Data Privacy Exposures
Using AI tools to evaluate, formate or otherwise process sensitive data also creates potentially serious risks because the use of these tools typically involves allowing the AI tool to access or uploading the confidential or other sensitive information into the tool. Organizations should use policies and processes to prevent their people’s use of AI tools need to use care to avoid violating statutory, regulatory or contractual confidentiality requirements, compromising their own trade secrets, proprietary information and intellect property, or both.
As an initial matter, organizations should avoid using AI tools that require uploading customer, financial, sales, or other data or information that the organization considers its own trade secrets or proprietary information into AI data bases or tools that do not adequately safeguard the ownership and confidentiality of that information.
Uploading or sharing trade secrets, Health Insurance Portability and Accountability (“HIPAA”), confidential employee, tax or other regulated information, trade secrets or other confidential or sensitive data into AI tools or searches without proper controls may itself breach of HIPAA, trade secret, federal or state privacy laws (e.g., biometric, consumer data laws) or other statutory, regulatory, ethical or contractual data privacy or confidentiality obligations. Additionally, allowing AI tools to access and interact with electronic data or systems frequently triggers data and systems security obligations under HIPAA, the Fair and Accurate Credit Transactions Act, electronic crimes, federal and state government contract, and a broad range of international, federal and state cybersecurity laws and regulations, and other government and private contractual and program participation, statutes, and regulations.
Concurrently, not all data and information obtained through a ChatGPT or other AI search is shareware. Most nongovernmental data bases, contractual firms and templates and other materials accessed through AI searches are or incorporate are owned or subject to copyrights or other intellectual property protections owned by third parties. AI searches can and often do access and incorporate data and other resources protected by third party copyright,trade secrets, HIPAA or other confidentiality, or other safeguards. Accordingly, accessing or using data bases, sample language or forms, or other materials without proper licensing or attribution may trigger liability to individuals and organizations for breaches of these intellectual property rights.
AI Tool Hallucinations and Other Output Deficiency Risks
AI tools and the output they produce are not always reliable. Among other things, certain AI tools are known to:
- Lack the ability to distinguish between more and less credible information sources;
- Create plausible-sounding but entirely fabricated facts, news articles, legal authorities, or academic citations;
- Create biased, false, incomplete, or inaccurate responses when models lack complete training data, are subjected to biased data, have limited context, ir under other circumstances;
- Create false positives such as I dentifying a threat (e.g., in fraud detection) that is not actually present.
- Fail to detect a real threat (e.g., in medical imagery) or report other false positives;
- Fabricate non-existent, fake information or other incorrect or inaccurate information; and
- Engage in hallucinations or other errors
The quality of the response, at best, often varies based on the quality and precision of the question asked. Lack of experience and careful structuring of the questions and inquiries made, lack of specialized knowledge necessary to structure the inquiry to be tailored to the specific needs at hand, and other limitations and concerns about the searches can undermine the accuracy, completeness and relevance of the AI tools output. Accordingly, response obtained by AI tools, often are unreliable and must be validated by a person experienced and skilled. The validation process should be conducted in such a matter that it preserves evidence that changes and responses are made based on thoughtful and reasonable determinations that the evidence obtained was not applicable or reliable, to minimize susceptibility to claims that decisions and actions were cherry picking based on improper intent rather than appropriate quality assurance processes. Organizations allowing the use of the tools and the individuals utilizing them need to understand and appropriately manage the very operational, legal in other risks of these deficiencies and error errors when utilizing AI tools.
Adopt And Enforce AI Policies To Manage AI Tool Use Responsibilities and Risks
Considering these and other responsibilities, organizations need to take thoughtful steps to decide when, how, why, and by whom it allows AI tools to be used in or on behalf of its organization and provide appropriate steps to manage those uses in the resulting ESI to fill its legal obligations and manage its legal and operational risks. Because this process of itself could be evidence impacting, the organizations, legal exposures, organizations generally should work with qualified legal counsel within the scope of attorney. Client privilege to work out these issues, define policies and practices, and operationalize those into an optimal process to promote the organizations legal and operational interests.
The author of this update, Cynthia Marcotte Stamer has decades of experience advising and representing governmental and private entities, AI and other technology, workforce and other legal and operational compliance, risk management and other operational and enforcement matters. If you have questions or need advice or help evaluating or addressing these or other compliance, risk management, or other concerns, contact her.
For More Information
We hope this update is helpful. For more information about these or other legal, contractual or operational compliance or risk management, please contact the author Cynthia Marcotte Stamer via e-mail or via telephone at (214) 452-8297.
Solutions Law Press, Inc. invites you receive future updates by registering on our Solutions Law Press, Inc. Website and participating and contributing to the discussions in our Solutions Law Press, Inc. LinkedIn SLP Health Care Risk Management & Operations Group, HR & Benefits Update Compliance Group, and/or Coalition for Responsible Health Care Policy.
About the Author
Recognized by her peers as a Martindale-Hubble “AV-Preeminent” (Top 1%) and “Top Rated Lawyer” with special recognition LexisNexis® Martindale-Hubbell® as “LEGAL LEADER™ Texas Top Rated Lawyer” in Health Care Law and Labor and Employment Law; as among the “Best Lawyers In Dallas” for her work in the fields of “Labor & Employment,” “Tax: ERISA & Employee Benefits,” “Health Care” and “Business and Commercial Law” by D Magazine, Cynthia Marcotte Stamer is a practicing attorney board certified in labor and employment law by the Texas Board of Legal Specialization and management consultant, author, public policy advocate and lecturer widely known for her more than 35 years of health industry and other management work, public policy leadership and advocacy, coaching, teachings, and publications including leading edge work on workforce and other risk management and compliance.
Ms. Stamer’s work throughout her career has focused heavily on working with businesses domestically and internationally on employment, benefits, technology, data confidentiality, privacy, and security, and other Federal Sentencing Guidelines and other workforce management, regulatory and public policy and other legal and operational concerns.
Author of many highly regarded compliance, training and other resources on these and other operations, risk management, compliance and government affairs concerns, Ms. Stamer is widely recognized for her thought leadership and advocacy on these matters.
In addition, Ms. Stamer currently or previously served as the American Bar Association (“ABA”) Joint Committee on Employee Benefits OCR annual agency scribe and a Council Representative, International Section International Employment Law Committee Chair and International Life Sciences and Health Committee Chair, ABA TIPS Medicine and Law Committee Chair, ABA Health Law Section Managed Care & Insurance Interest Group Chair, former Vice President and Executive Director of the North Texas Health Care Compliance Professionals Association, past Board President of Richardson Development Center (now Warren Center) for Children Early Childhood Intervention Agency, past North Texas United Way Long Range Planning Committee Member, and past Board Member and Compliance Chair of the National Kidney Foundation of North Texas, and a host of other professional and civic leadership roles. She is a Fellow in the American College of Employee Benefit Counsel, the American Bar Foundation and the Texas Bar Foundation, Ms. Stamer also shares her extensive publications and thought leadership as well as leadership involvement in a broad range of other professional and civic organizations.
For more information about Ms. Stamer or her health industry and other experience and involvements, see www.cynthiastamer.com or contact Ms. Stamer via telephone at (214) 452-8297 or via e-mail here.
About Solutions Law Press™
Solutions Law Press™ provides health care, human resources and employee benefit and other business risk management, legal compliance, management effectiveness and other coaching, tools and other resources, training and education on health care, leadership, governance, human resources, employee benefits, data security and privacy, insurance, and other key compliance, risk management, internal controls and operational concerns. If you find this of interest, you also be interested reviewing some of our other Solutions Law Press™ resources.
If you or someone else you know would like to receive future updates about developments on these and other concerns, please be sure that we have your current contact information including your preferred e-mail by creating your profile here.
NOTICE: These statements and materials are for general information and purposes only. They do not establish an attorney-client relationship, are not legal advice or an offer or commitment to provide legal advice, and do not serve as a substitute for legal advice. Readers are urged to engage competent legal counsel for consultation and representation considering the specific facts and circumstances presented in their unique circumstance at the particular time. No comment or statement in this publication is to be construed as legal advice or admission. The author reserves the right to qualify or retract any of these statements at any time. Likewise, the content is not tailored to any particular situation and does not necessarily address all relevant issues. Because the law constantly and often evolves, subsequent developments that could impact the currency and completeness of this discussion are likely. The author and Solutions Law Press, Inc. disclaim and have no responsibility to provide any update or otherwise notify anyone of any fact or law specific nuance, change, limitation, or other condition that might affect the suitability of reliance upon these materials or information otherwise conveyed in connection with this program. Readers may not rely upon, are solely responsible for, and assume the risk and all liabilities resulting from their use of this publication.
Circular 230 Compliance. The following disclaimer is included to ensure that we comply with U.S. Treasury Department Regulations. Any statements contained herein are not intended or written by the writer to be used, and nothing contained herein can be used by you or any other person, for the purpose of (1) avoiding penalties that may be imposed under federal tax law, or (2) promoting, marketing or recommending to another party any tax-related transaction or matter addressed herein.
©2026 Cynthia Marcotte Stamer. Non-exclusive right to republish granted to Solutions Law Press, Inc.™ For information about republication, please contact the author directly. All other rights reserved.
Cynthia Marcotte Stamer
slproc.com 
You must be logged in to post a comment.