This article examines some of the key legal risks associated with AI note-takers and offers some practical governance considerations for those organisations looking to leverage the technology.

You can also read IP/ICT Senior Associate, Michelle Dunlop’s interview with the NBR tech editor on this topic here [paywall].

What is an AI note-taker?

AI note-takers are a form of machine learning technology that automate the minute-taking process by transcribing spoken words into text. The AI generates summaries of key discussion points, tracks action items, and provides meeting insights.

Popular AI note-taking tools, such as Otter.ai and Fireflies, are increasingly being used at board and executive leadership meetings across various industries. The appeal lies in the AI tool’s ability to automate an otherwise administrative process, freeing up time for the leadership teams to focus on strategic discussions.

AI note-takers are typically used to augment traditional transcribing functions. All required participants attend the meeting, and the tool is activated at the outset to record the meeting. However, some AI note-taking tools can join meetings in place of (and at the request of) an absent individual, prompting questions about whether a director could send an AI to attend a board meeting in that director’s place.

Does this mean we could have a 'Robot Director' and can AI perform a director’s functions?

No. Under the Companies Act 1993, directors must be natural persons, meaning that (currently) it is not possible to appoint an AI as a director. In any event, an AI note-taker is merely transcribing what occurred at the relevant meeting.

While it may be tempting for a busy director to send an AI to attend a board meeting in his or her place (but with a view only to the busy director being informed as to what took place at the meeting), a director should bear in mind that an AI note-taker would not count toward quorum requirements, meaning a meeting may fail to meet the required thresholds for decision-making purposes.

More importantly, the AI note-taker cannot fulfil the role expected of a director. Directors cannot delegate their fiduciary duties and using AI will not absolve a director of that director’s legal responsibilities.

What should we consider before inviting an AI to our meeting?

• Accuracy: While AI note-taking tools are improving at a rapid rate, they are still prone to hallucinations. The AI may generate inaccurate quotes or extrapolate beyond what was actually said. The AI may not fully understand nuanced conversations or technical language, and may struggle with accents, colloquialisms, and other speech nuances. These factors could lead to the AI producing incorrect or inaccurate meeting transcripts and relying on these could put a director at risk of breaching the director’s statutory duties of care and diligence.
• Data security: AI note-taking tools record data from the meeting, some of which may be highly commercially sensitive. That data may be subsequently transferred and stored on the systems of the tool’s provider, with no easy way for a company to delete it. Those systems could be vulnerable to cyberattacks. In addition, some vendors use collected information for the further training and iterations of their AI models, risking a company’s confidential information being inadvertently exposed to third parties.
• Privacy: AI note-takers record voices, and content that could include the personal news or personal information about participants, raising consent and privacy issues, particularly if meeting participants are unaware of the recording.
• Confidentiality and privilege: AI note-takers are programmed to automatically disseminate outputs following the conclusion of a meeting. If AI note-takers capture legally privileged or commercially sensitive information, that information could end up being disseminated more widely than intended and risk privilege and/or confidentiality being waived in the relevant materials.
• Litigation risks: AI-generated minutes could create multiple records that subsequently become discoverable as part of legal proceedings. Complications could arise from an evidentiary perspective if the relevant records are unvetted, unverified, and contain discrepancies.
• Public sector considerations: For meetings involving public sector organisations, AI-generated records could become disclosable under the Official Information Act or similar legislation.

How can we mitigate the risks?

• Have a clear policy on the use of AI note-takers. For board and executive team meetings, AI note-takers should generally only be used when participants are present and with the consent of all other members. It may be that the content to be discussed at a meeting is considered too sensitive for AI to be utilised at all.
• Select AI tools from reputable vendors; ideally a tool that allows meeting participants some degree of control over its functionalities, such as disabling recording when necessary.
• Consider security and data privacy requirements. This includes reviewing the tool’s privacy policies to determine where data is stored, how it is handled, and whether the vendor retains rights to use collected data for further training.
• Human oversight is key to maintaining the accuracy and integrity of AI-generated records. Promptly review and correct AI-generated outputs, confirm the official meeting record, and delete transcripts when they are no longer needed.

Get in touch

If you would like to discuss any aspect of this article, or need any assistance, please contact one of our experts

Special thanks to James Burnett for his assistance in writing this article.