AI notetakers have moved beyond meeting bots. They now run on smartwatches, AI pendants and smart glasses, and they can record conversations with no visible signal. In this CSO Online article, Joe Sullivan, founder of Joe Sullivan Security and former CSO of Uber, Cloudflare and Facebook, and Gadi Evron explain why most company policies, written for bots that visibly join video calls, no longer cover the security, privacy and legal risks of AI recording. They look at how transcripts can be manipulated through prompt injection, how one misheard word can change a whole summary, and how tools that draw on past meetings can leak internal context to outside parties.

The authors give security, legal and HR leaders six steps for governing AI notetakers: require disclosure whenever a recording tool is running, list in writing the meetings that can never be recorded (legal, HR, M&A and board sessions), track when outside parties record your employees, set your own transcript retention rules instead of relying on vendor defaults, review each vendor’s security and data reuse terms, and assign a named person to verify any transcript that becomes an official record.

Read the full article on CSO Online →