Artificial Intelligence in Practice: Is Your AI Notetaker Helping You or Hurting You?
Category | Briefing Papers
By Leon P. Wells and law clerk Cassidy P. Hartig
Leon is a shareholder of the firm’s Construction Law Department. He can be reached at 612.359.7651 or LWells@fwhtlaw.com
As regular readers of FWHT briefing papers will know, the use of generative AI across the construction industry presents not only exciting opportunities for project-related efficiencies and project management, but also legal risks when it comes to project-related litigation. As our colleague Jesse Orman wrote this past July in his article Your AI Chat Logs May Be Exhibit A, the benefits of AI use in the construction industry are obvious:
A project executive can upload thousands of pages of project documentation and ask an AI system to identify potential change-order issues. A project manager can ask an AI tool to evaluate whether a subcontractor’s notice complies with contractual requirements. A claims consultant can use an AI platform to summarize competing schedule analyses. And when disputes arise, personnel often use generative AI to organize facts, test arguments, and evaluate potential claims before consulting counsel. What not so long ago felt like an experimental technology has become an increasingly common business tool.
Jesse then pointed out some of the clear project-related risks that active use of AI tools can present:
In each instance, [AI] prompts themselves may reveal the user’s thinking. The uploaded materials may identify what facts the user considered important. The AI output may contain favorable or unfavorable analyses. Follow-up questions may expose perceived weaknesses in a claim or defense. The AI chat history could very well appear to be a roadmap of the organization’s internal evaluation process. Experienced construction litigators immediately recognize the significance. Opposing counsel routinely seek documents revealing contemporaneous assessments of project issues. A claim analysis generated through an AI platform may prove every bit as interesting to an adversary as an internal memorandum prepared by a project executive.
The takeaway is clear: Be careful about the information you choose to put into AI tools and what you actively ask the AI tool to do with the information. In the event of litigation, a user’s prompts, uploads, follow-up questions, and general AI chat history may provide outsiders the ability to “pull back the curtain” on contemporaneous project assessments in a manner that was not possible just a few years ago.
But what about those AI tools that are not used in the same active, intentional way? As AI continues its march toward maximizing efficiency, many AI tools have been developed to stay in the background, more or less unnoticed. These tools include AI notetakers, frequently used by businesses to automatically and unobtrusively listen in on calls and meetings, analyze conversations, take meeting notes, and email post-meeting summaries to meeting participants and others.
Like other AI tools, the appeal of AI notetakers is undeniable. In today’s world, businesses—particularly, stakeholders in large, complex construction projects—face an almost overwhelming swell of calls and virtual meetings. But we all know that meetings and calls aren’t just limited to discussion of ongoing projects. Virtual meetings are increasingly used in hiring; firing; other HR-related meetings (like employee performance reviews); incident investigations; lessons-learned meetings; other business-related calls; and even, sometimes, calls with attorneys. At the same time, businesses and their employees face ever-growing pressure not only to keep better records, but to keep them more efficiently. AI notetakers therefore offer an attractive option for reducing this burden and, in many workplaces, have become critical tools on which businesses rely for organization and recordkeeping.
However useful AI notetakers may be, their use can present particular risks of which businesses should be aware—from creating new liability exposure to eviscerating confidentiality and privilege protections a business may otherwise enjoy when speaking internally or with external counsel. This article examines this unique AI tool by briefly exploring different examples of the tool; basic differences in how these tools operate; how the use of an AI notetaker can go wrong; and simple steps businesses can take to protect themselves from the risks posed by the deployment of AI notetakers.
- What are AI Notetakers?
For the uninitiated, think of AI notetakers as a robotic secretary or stenographer that listens in on a virtual meeting or video call, analyzes the discussion in real time, and creates transcriptions or summaries for meeting participants and others. Examples include stand-alone tools like Read AI, Granola, Fathom, Otter, and Fireflies, as well as more integrated options like Microsoft’s Copilot (if transcription and summarization settings are enabled).
Sometimes, the AI notetaker enters the call or meeting as separate participant (or “bot”)—one that shows up on the attendee list and has its own window on the virtual meeting platform. In these cases, every meeting participant can see that the AI notetaker is present, and the AI notetaker is usually easily identifiable by name. However, bot-free AI notetakers are becoming more and more common. These versions of the tool listen in to meetings from users’ devices without adding a visible bot to the call and without necessarily alerting the meeting participants to their presence. In either case, the AI notetaker listens to conversations, processes spoken language, and—depending on setup—will either produce a verbatim (or near verbatim) transcription, or it will attempt to analyze the conversation and its context to produce meeting notes or summaries. AI notetakers can then email the transcription or summary to meeting participants, invitees, or others.
Needless to say, the quality of a transcription or summary can vary. The AI notetaker may produce a spot-on transcription or correctly understand the context of the meeting such that it can produce an accurate summary of who said what or list of action items, key takeaways, or conclusions reached during the meeting. But accuracy is not guaranteed. As is likely noted somewhere in the AI notetaker’s product literature, “AI can make mistakes”—including misunderstanding speech, misattributing one person’s speech to someone else, or misunderstanding the overall context or purpose of the meeting (particularly when the context isn’t expressly stated during the meeting).
Humans make mistakes, too. Incorrect settings may allow the AI notetaker to share post-meeting notes more broadly than intended. The meeting organizer may accidentally allow an AI notetaker to listen in instead of turning the notetaker off. The AI notetaker may be configured to allow—or the business’s license for the AI notetaker may allow—the broader use of recorded information for training the AI or improving general user experience with the tool.
Each of these mistakes could have unintended consequences about which businesses need to be aware.
- An Example of AI-Notetaker Use Gone Wrong
Although the potential consequences of mistakenly using AI notetakers are too varied to address in a single briefing paper, a recent lawsuit in New Jersey highlights the risks of automatically using AI notetakers in every meeting and configuring AI notetakers to automatically send out meeting notes to all participants.
In Waninger v. Marathon Engineering & Environmental Services Inc.—a lawsuit filed in New Jersey state court in August 2026—a former employee of Marathon Engineering sued the company alleging gender discrimination. According to the employee’s complaint, an AI notetaker was permitted to listen in on her virtual termination meeting, record a manager’s post-termination comments (made after the terminated employee had left the meeting but before the management team ended the meeting), and email a transcript of the meeting to all meeting participants. The post-termination comments by management allegedly included statements about “what our ideal person looks like” for replacing the female employee: “[h]opefully a relatively strapping young man.” Because the AI notetaker heard these comments and was configured to send a transcript of the meeting to all participants, the just-terminated employee received a meeting transcript, including management’s comments, by email.
Although the Waninger lawsuit is just getting underway and the employee’s allegations have yet to be proven, the import of management’s alleged comments, if true, is obvious: the decision to fire the employee may have been—at least in part—based on the employee’s gender, leading to potential legal liability.
- Other Disclosure Risks Posed by AI Notetakers
The circumstances presented in Waninger are extreme, and the case provides a dramatic example of how the use of AI notetakers, if not carefully managed, can have unintended legal consequences for businesses. But one can easily imagine less extreme examples.
Imagine if in-house counsel had accompanied the management team in the termination meeting described in the Waninger case. Even if management’s alleged comments never occurred, management and counsel likely would have debriefed while the AI notetaker was still listening—discussing how they thought the termination went, what went right, what went wrong, what risks arose or were avoided during the termination meeting, and so on. Counsel and the management team would certainly have wanted and expected this type of discussion to remain confidential; but because the AI notetaker was configured to automatically send meeting notes to all meeting participants (including the terminated employee) this confidentiality would have been quickly destroyed. Not only could a terminated employee could easily pick through these discussions and, rightly or wrongly, use them as ammunition in a wrongful termination suit against the company, any attorney-client privilege that would have attached to the discussion between counsel and management would likely be destroyed.
That said, the mere use of an AI notetaker—even if properly configured—can present risks to attorney-client privilege, too. Although the caselaw around AI tools and traditional rules regarding discoverability, privilege, confidentiality, and work product is still emerging and very much in flux around the country—as discussed previously by our colleague Jesse Orman—Minnesota’s existing caselaw may provide some hints on how courts may rule when faced with claims that the use of AI notetakers destroys confidentiality.
For example, in In re State of Minnesota v. Flowers, 986 N.W.2d 686 (Minn. 2023), the Minnesota Supreme Court considered whether an incarcerated person who knowingly called his attorney on a recorded line instead of the government-provided confidential one could invoke attorney-client privilege. The Court determined the defendant had waived his privilege claim because he knew the recorded calls were monitored and ignored the private option. Flowers presents an instance where a client lost attorney-client privilege because they could have maintained privacy but did not. As the Minnesota Supreme Court has previously noted, whether materials or discussions are protected by the attorney-client privilege may turn on “whether the client intended to keep the specific [information] confidential and whether the client and attorney took all steps reasonably necessary to prevent disclosure.” Kobluk v. Univ. of Minnesota 574 N.W.2d 436, 444 (Minn. 1998) (emphasis added). In Flowers, the client obviously did not take “all steps reasonably necessary to prevent disclosure.”
Based on this precedent, one could foresee a court determining that AI notetaking is not covered by the attorney-client privilege when the user agrees to terms of use that specifically allow the provider to use or sell data—as is often permitted for purposes of improving AI models and other users’ experiences. In such circumstances, a court may find a failure to take “all steps reasonably necessary to prevent disclosure.” On the other hand, if a client uses an AI product that promises to shield data from third parties, courts may see this as a reasonably necessary step toward privacy and one sufficient to maintain privilege. That said, legal commentators are not in full agreement. Some maintain that privilege may be destroyed regardless of the terms of use because even the paid version that offer greater privacy protections allow “AI companies [to] still hear what is said in [the] room” between attorney and client. Vincent J. Syracuse et. al., Talking to AI Could Talk You Out of Attorney-Client Privilege, 98-SUM N.Y. St. B.J. 56, 57 (Summer 2026).
- How can businesses protect themselves as the court grapples with AI Notetakers?
Although it may be impossible to eliminate all risks associated with AI notetakers, a few key practices may go a long way in offering protection from the various legal risks they pose:
- • Configure AI notetakers so that they do not automatically join calls. When a user is forced to consider whether or not the use of the notetaker is needed or desirable, the user is less likely to introduce the notetaker to particularly sensitive meetings.
- • Configure AI notetakers so that distribution of notes is tightly controlled. Notes should go first to a point-person who can review the substance and accuracy of the notes.
- • Eliminate AI notetakers, as a matter of policy, from HR meetings or meetings and calls with lawyers. Utilizing ordinary and pre-approved paperwork when it comes to HR-related meetings will maintain a cleaner record to which AI notetakers are unlikely to add much value. Eliminating AI notetakers from calls with lawyers is the best way to ensure you don’t breach the attorney-client privilege with an AI notetaker.
- • When an AI notetaker is being used, announce its involvement to the meeting participants. Although not specifically discussed above, this can help reduce the risk of others making claims that they were wrongfully or illegally recorded under privacy or wiretapping laws.
- • Request all parties in a confidential meeting to turn off any bot-free AI Notetaking tools that may not be visible in the meeting interface. This can help protect businesses from unknowingly having meetings recorded by third parties or employees.
- • Create company policies and educate employees on the risks of AI Notetaking—and AI generally—to protect confidential business information. Education and intentional use is one of the best ways to reduce the risks posed by careless use of AI tools. Among other things the policies should consider what AI tools are acceptable, who can use AI tools, and under what circumstance the use of AI tools is permitted.
Like all new technology, AI presents inherent risks—especially as the legal system tries to catch up with the technology. Staying up to date on best business and litigation practices when it comes to AI can be the difference between AI being a helpful assistant or a dangerous adversary.
Announcements
Congratulations to Fabyanske’s Westra, Hart & Thomson’s P.A.’s Julia Douglass for being elected as a new Fellow of the American College of Construction Lawyers
On September 3rd, Julia was announced as an elected Fellow by the ACCL board.
Fellowship is extended by invitation to those who are found to have mastered the practice or the teaching of construction law and dispute resolution in the complex technical and legal fields pertaining to the built environment, whose professional careers have been marked by the highest standards of ethical conduct, scholarship, professionalism, and collegiality, and who have demonstrated a commitment to “give back” to the construction industry.
Fabyanske, Westra, Hart & Thomson, P.A. is pleased to announce the following thirteen attorneys have been selected as Best Lawyers by their peers in the recent Best Lawyers© publication, one of the oldest and most respected peer-review publications in the legal profession: Scott Anderson, Mark Becker, Hugh Brown, Matt Collins, Julia Douglass, Rory Duggan, Gary Eidson, Kyle Hart, Jeffrey Jones, Jesse Orman, Robert Smith, Dean Thomson, and Katie Welsch. Also, congratulations to Rob Smith for being named “Lawyer of the Year” by Best Lawyers® for Litigation – Construction, Minneapolis (2027) and to Katie Welsch for being named “Lawyer of the Year” by Best Lawyers® for Banking and Finance Law, Minneapolis (2027)
Congratulations to Fabyanske, Westra, Hart & Thomson, P.A. attorneys who have been named The Best Lawyers in America: Ones to Watch (2027 Edition). They are Alexander Athmann and Colin Bruns.