Instinct AI Sparks Privacy Fears Over Perpetual Data Rights and Binding User Transactions
Updated
Updated · TechCrunch · Aug 24
Instinct AI Sparks Privacy Fears Over Perpetual Data Rights and Binding User Transactions
1 articles · Updated · TechCrunch · Aug 24
Summary
Private testers say Instinct’s standout automation comes with unusually broad access to email, messages, calendars, location, audio, screen activity and keyboard inputs, raising fresh privacy and security alarms.
Terms circulated online grant the assistant a “perpetual and irrevocable” license to access, store, modify and use user materials for AI training, and let it enter agreements or transactions on a user’s behalf.
Early incidents sharpened those concerns: Peter Yang said Instinct initially would not delete his Gmail records, while Claire Vo said it kept summarizing her inbox after disconnection and stored emails in plain text.
Other testers said the agent could pull sign-up codes from inboxes to complete tasks, was vulnerable to phishing, and in one case sent an email without prior approval, undermining trust in autonomous actions.
Instinct remains in private access and has not publicly addressed the complaints; the San Francisco startup, led by former Sierra scientist Noah Shinn, is reportedly backed by Kleiner Perkins and Conviction.