Employers are seeing a marked increase in the use of generative artificial intelligence (“AI”) platforms by employees and former employees. It is becoming quite common for an employer to receive a detailed demand letter asserting employment related claims, written not by an attorney, but by a former employee with the assistance of ChatGPT, Claude, or a similar AI platform. But what about use of AI platforms by Human Resources? There is no doubt that AI can help employers streamline certain personnel processes and procedures, making tasks related to recruiting, hiring, training and even performance more efficient. There is, however, one important caveat that employers need to keep in mind: documents created using AI are generally discoverable in a lawsuit unless they are generated at the request of your attorney in anticipation of litigation or for trial, and, even then, there are limitations.
Consider the following examples:
- Example 1: You are concerned that you may have certain employees misclassified as exempt under the Fair Labor Standards Act (“FLSA”). You enter a description of the duties these employees perform into one of the publicly available AI platforms and ask questions related to whether the employees are exempt from overtime. The responses indicate that certain employees do not meet the FSLA’s exempt duties test and should instead be classified as non-exempt, entitling them to overtime pay for any hours worked over 40 in a workweek. You email the information to the CFO and a co-worker in HR.
- Example 2: You have received several complaints regarding a particular manager during the last six months, including complaints regarding sexual comments the manager has made to female employees. You take the information from your notes on discussions with employees who have raised concerns about the manager and ask an AI platform to create a summary. You put the summary in a file marked “investigation.”
In both of the above scenarios, assuming there were no discussions with the company’s attorney in which the attorney asked for the documentation created using AI in anticipation of litigation, you have created materials that are not only discoverable in litigation, but damaging to the company’s position. In the first example, an employee (or former employee) who sues for back wages in the form of overtime pay because he or she should have been classified as non-exempt, can argue that as of the date of the AI generated response, the company had knowledge that it was violating the FLSA, effectively eliminating any argument that the company acted in good faith or had a reasonable basis for believing it was in compliance with the FLSA.
In the second example, the summary of “bad acts” by the manager that was created using AI may appear more damaging than the individual notes from your discussions with employees. The attorney for the employee or former employee alleging sexual harassment by the manager in violation of Title VII of the Civil Rights Act has a document that neatly summarizes all the complaints and can be used to show the company had knowledge of unlawful harassment and failed to prevent it from reoccurring.
The Heppner Case
The perils of using AI to address or investigate potential claims was recently analyzed by the United States District Court for the Southern District of New York. On February 17, 2026, the Court looked at what appeared to be a question of first impression in the federal courts: “when a user communicates with a publicly available AI platform in connection with a pending criminal investigation, are the AI user’s communications protected by the attorney-client privilege or the work product doctrine?”1 Although this case involved a criminal defendant, the court’s reasoning is informative when considering whether AI-generated documents are discoverable in litigation involving civil claims, including employment-related claims. In Heppner, the defendant, an executive of several entities, was charged with securities fraud, wire fraud, and falsifying corporate records, among other charges. In connection with Heppner’s arrest, the FBI seized thirty-one documents memorializing communications Heppner had with the generative AI platform, Claude. These communications included an outline of arguments he could make regarding the facts and law related to the potential charges against him.
Heppner’s lawyer argued that the AI documents reflecting Heppner’s communications with Claude, which took place after it was clear he was subject to an FBI investigation, were protected by both the attorney-client privilege and the work product doctrine and, therefore, were not discoverable by the U.S. Government.2 According to Heppner’s lawyer, Heppner created the AI documents for the purpose of speaking with his attorney to obtain legal advice and subsequently did share the documents with his legal counsel. He conceded, however, that he did not direct Heppner to run Claude searches.
The court ruled that Heppner’s AI-generated documents were not protected by the attorney- client privilege or the work product doctrine.3 With regard to the attorney-client privilege, the AI documents did not reflect communications between Heppner and his attorney, since Claude is not an attorney. In addition, the court reasoned that the communications memorialized in the AI documents were not confidential. Claude is a third-party AI platform with a privacy policy, to which the user consents, which allows the collection of data and grants the right to disclose such data. The AI documents at issue were “not like confidential notes that a client prepares with the intent of sharing them with an attorney because Heppner first shared the equivalent of his notes with a third-party, Claude.”4 The court noted that non-privileged communications do not become privileged ones simply by sharing them with legal counsel. Notably, even if the information that Heppner provided to Claude had included privileged discussions with his attorney, Heppner waived the privilege by disclosing it to a third party, Claude.5
As for the work product doctrine, the court found that the AI documents were not protected by this doctrine because they were not prepared at the request of counsel, nor did they reflect defense counsel’s strategy.6 There was no dispute that Heppner acted on his own when he created the AI documents. Because the AI documents “were prepared by the defendant on his own volition,” he was not acting as his counsel’s agent when he communicated with Claude.7
What Does this Mean for Employers?
HR professionals and others who are investigating or researching personnel issues need to keep in mind that communications with public AI platforms will generally be discoverable in litigation except in very limited situations in which the attorney-client privilege or work product doctrine applies. By consulting with an AI platform, you may be creating documentation that will later adversely affect the employer’s position if litigation ensues. Companies and Human Resources departments should maintain their own documentation and seek legal advice when needed, but think before communicating with an AI platform or having AI create documentation related to personnel issues.
Need some Insider help? Contact Cherie Blackburn or Christy Rogers.
1 United States v. Heppner, 820 F. Supp. 3d 292, 294 (S.D.N.Y. 2026).
2 The attorney-client privilege applies to confidential communications between an attorney and his or her client for the purpose of obtaining or providing legal advice. The work product doctrine generally protects documents and tangible things that are prepared in anticipation of litigation or for trial from discovery by the opposing party. Rule 26(b)(3), Fed. R. Civ. P. This doctrine, “[a]t its core[,] . . . shelters the mental processes of the attorney, providing a privileged area within which [the attorney] can analyze and prepare [the] client’s case.” Id., at 297, citing, United States v. Nobles, 422 U.S. 225, 238, (1975). The work product doctrine can include materials prepared by the attorney or persons other than the attorney, provided the materials were created to prepare for litigation.
3 Heppner, at 299.
4 Id., at 297.
5 Id., at 297, .n. 3.
6 Id., at 298, citing, Matter of Grand Jury Subpoenas, 959 F.2d 1158, 1167 (2d Cir. 1992).
77 Id., at 298
