AI notetakers have become a fixture of modern meetings, promising to transcribe conversations, summarise discussions and generate to-do lists without anyone having to type a word. But as adoption grows across boardrooms and client calls, a growing number of professionals, lawyers and privacy advocates are raising concerns about exactly what these tools quietly collect, and where that data ends up.
More Than Just a Transcript
An AI notetaker uses speech recognition and large language models to record, transcribe and summarise conversations in real time. What many users do not realise is that most of these tools also generate a voiceprint, a biometric profile unique to each speaker’s vocal characteristics, similar in principle to a fingerprint.
According to Chris Pluymers, an associate attorney at The Dillon Law Group, companies use these voiceprints to distinguish between speakers in a transcript, labelling them “Speaker 1” or “Speaker 2.” The concern, he said, is that voiceprints are also used elsewhere to verify identity, including for accessing bank accounts over the phone, meaning a leaked or misused voiceprint could theoretically be exploited for fraud or unauthorised account access.
Privacy advocates have also flagged that some technology companies behind these tools resell collected data or use confidential meeting recordings and transcripts to train their own AI models, often without participants being fully aware of how their voice data will be reused down the line.
The Consent Blind Spot
A particular concern among professionals is that AI notetakers are frequently deployed by just one participant using a personal device, separate from the official meeting platform, meaning other attendees may have no idea a conversation is being recorded and transcribed at all. Technology columnist Thorin Klosowski has suggested that the more considerate approach is for the person deploying the tool to disclose it upfront, rather than leaving other participants to assume.
Some professionals have adopted workarounds to manage the risk. Employment consultant Amy Dufrane recommends that companies establish a blanket policy stating that certain meetings cannot be recorded, which relieves individual employees of having to raise objections themselves. Others simply switch off the notetaker partway through a meeting once the conversation turns to sensitive matters, preferring to lose the transcript rather than risk a confidential exchange being logged.
In parts of the United States, this is no longer just a matter of workplace etiquette. Illinois law treats voiceprints as biometric identifiers on par with fingerprints, requiring written notice, informed consent, and a documented data retention and destruction policy before any AI notetaker can capture them, protections Pluymers noted most companies deploying these tools have yet to put in place.
Where India’s Own Rules Stand
India does not yet have an equivalent operative framework, but the groundwork is being laid. Under the Digital Personal Data Protection Act, 2023, an individual’s voice and biometric characteristics fall squarely within the definition of personal data, and any AI tool that captures, stores or processes that data without proper consent carries meaningful regulatory risk once the Act is fully enforced. The law requires that consent be free, specific, informed and revocable through a clear affirmative action, meaning a generic disclaimer at the start of a call is unlikely to satisfy the standard for something as sensitive as a voice recording.
The DPDP Act’s core provisions are set for full enforcement by May 2027, and penalties for non-compliance can run up to ₹250 crore per violation. Until then, the older Information Technology Act and its associated data rules continue to govern how Indian companies handle recorded voice data, leaving a regulatory gap that mirrors the uncertainty many professionals abroad are already navigating in their day-to-day meetings.
For now, the simplest safeguard available to most professionals, in India or elsewhere, remains asking a direct question before a meeting begins: is this conversation being recorded, and by what.
