You’re not sending an email. You’re training a model

AI meeting note-takers look like productivity tools, but their transcripts are also permanent training datasets that can score contribution and coaching needs. The stakes are illustrated by Google’s proposed $10m purchase of Spirit Airlines’ bankruptcy data—including about 100m emails and 500m Teams messages—that could teach aviation pricing and fuel costs plus negotiation and strategy rarely available online. Rival bidder Micro1, software firm Springshot and a Spirit flight-attendants’ union challenge the deal over confidential employee data; a court hearing is due October 14th. Oxford’s Jeremias Adams-Prassl notes that if Spirit were European, GDPR purpose limitation would block simply reusing operational data to train AI. European privacy expectations also follow the 2017 Barbulescu ruling that employers cannot erase private social life at work. More permissive US rules may help frontier labs acquire proprietary corpora, but transcripts expand litigation discovery and worker resistance (e.g. Meta backlash over click/keystroke logging). If staff believe internal messages will train models that outlive the firm, they will write more guardedly—and workplaces may grow more professional and duller.