AI assistants in virtual meetings are recording and transcribing private discussions and jokes, sometimes with embarrassing results for users. These tools can automatically distribute summaries of entire calls, including informal small talk, to all participants. As reported by Ann-Marie Alcántara in the Wall Street Journal, this raises new concerns about privacy and context in the workplace.
The report highlights several incidents. A marketing agency owner was mortified when her new client received a transcript of her joking that he might be a “Nigerian prince” before he had joined the call. In another case, Zoom’s AI assistant created a meeting summary about a couple’s lunch plans, including that they discussed “the possibility of getting sandwich ingredients” and their dislike of soup.
While some teams have embraced the AI-generated bloopers as “comic relief,” the technology is changing user behavior. John Barentine, an astronomer, was surprised when a summary noted he had discussed the “lethal dose of water for humans.” His comment, made during a conversation about flooding, was stripped of its original context. He told the Journal he is now more likely to use private chat for remarks not intended for the entire group.
Companies like Zoom and Google emphasize user control. They state their tools provide clear visual and audio cues when transcription is active, and meeting hosts can disable the function at any time.