MIT researchers: agentic AI breaks privacy rules built on individual consent
_ghnn_ · x · 2026-11-23
MIT Sloan professor Catherine Tucker and PhD candidate Rozhina Ghanavi argue that privacy rules premised on individual disclosure break down with agentic AI.
- Example: a user sharing family medical history with an AI agent exposes relatives who never consented; the beneficiary of disclosure doesn't bear its full cost.
- Agents granted access to email, calendars, customer records and payments make retrieval/usage decisions on the user's behalf, making per-disclosure consent unworkable.
- Tucker says privacy rules will shape AI competition, and deletion requests leave open what trained models retain.
More from AGI Musings
- Infinite Computing: How AI and Quantum Computing Converge Into One Fabric — ChuckDBrooks · 2026-11-23
- Bindu Reddy: AI Is Reverse-Engineering the Code of a Simulated Universe — bindureddy · 2026-11-23
- Alexandr Wang: Riemann Hypothesis Proof Marks Superintelligence's Arrival — alexandr_wang · 2026-11-23
- DeepMind formalizes a 95-state machine that halts iff the Riemann Hypothesis is false — MikePFrank · 2026-11-23
- UK's 1860s railway mania seen as key to its late-1800s decline — _onionesque · 2026-11-23
- scaling01: Flood of new data in recent days has shortened my AGI timelines even further — scaling01 · 2026-11-23