[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-privacy-system-cuts-patient-data-leaks-in-medical-ai-to-near-zero":10,"sections":35},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},8872,"privacy-system-cuts-patient-data-leaks-in-medical-ai-to-near-zero","Privacy System Cuts Patient Data Leaks in Medical AI to Near Zero","A research prototype lets hospital AI agents tap remote models for help while cutting patient data leaks from 98% to 0.2% of cases.","A new AI system lets hospital chatbots phone a smarter, cloud-based AI for a second opinion without leaking who the patient is.\n\nResearchers built PrivMeSA, a system where a local AI agent runs the full patient encounter and only reaches out to a more capable remote model when it needs expert input. Instead of relying on a human-written rule to redact names or dates, the local agent uses reinforcement learning to weigh how much a detail helps accuracy against how much it risks exposing or re-identifying the patient, checking the entire outgoing message each time, not just the latest question. The system also keeps a running memory of past consultations, distilling them into general clinical lessons it can reuse on new cases without calling the remote model again. Tested on an emergency department benchmark built from real MIMIC-IV-ED hospital records, PrivMeSA boosted diagnostic accuracy by as much as 15.8 percentage points over a baseline delegation setup, while cutting direct disclosure of personal details from 98 percent of cases to 0.2 percent, and cutting cases where a patient could be narrowed down to ten or fewer people in a hospital registry from 74 percent to zero.\n\nThe underlying problem is real: hospitals want the better reasoning of large cloud models, but every patient detail sent off-site is a liability, and simply stripping names does not stop the accumulation of smaller clues such as age, diagnosis codes, and visit timing across multiple encounters. PrivMeSA's trick is treating privacy as something to actively manage across a whole conversation and across repeat visits, not a one-time redaction, and reusing past remote answers through memory instead of resending patient data every time.\n\nIt is one benchmark on emergency department data with simulated registry attacks, so the real test is whether 'self-evolving' memory still protects patients after thousands of messier, real-world cases.","[\"medical ai\",\"privacy\",\"multi-agent systems\",\"healthcare\"]","2026-10-01T04:00:00.000Z","2026-10-01T08:41:30.014Z","2026-10-01T08:41:36.305Z","published",null,[],"ai",[26,27,28,29],"medical ai","privacy","multi-agent systems","healthcare",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.38458",0,{"sections":36},[37,40,45,50,55,60,65,70,75,79,84,89,94,99],{"name":38,"slug":24,"count":39,"latest_published_at":18},"AI",5351,{"name":41,"slug":42,"count":43,"latest_published_at":44},"Security","security",801,"2026-09-30T22:18:23.000Z",{"name":46,"slug":47,"count":48,"latest_published_at":49},"Policy","policy",429,"2026-10-01T02:26:17.000Z",{"name":51,"slug":52,"count":53,"latest_published_at":54},"Deals","deals",298,"2026-09-30T21:00:26.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Hardware","hardware",196,"2026-09-30T13:00:00.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Science","science",157,"2026-09-30T15:00:56.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Consumer Tech","consumer-tech",149,"2026-09-30T22:57:11.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Dev Tools","dev-tools",93,"2026-10-01T02:30:48.000Z",{"name":76,"slug":77,"count":73,"latest_published_at":78},"Software","software","2026-09-30T21:41:11.000Z",{"name":80,"slug":81,"count":82,"latest_published_at":83},"Startups","startups",84,"2026-09-30T20:39:09.000Z",{"name":85,"slug":86,"count":87,"latest_published_at":88},"Gaming","gaming",51,"2026-09-30T16:24:30.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"General","general",50,"2026-09-30T21:37:54.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"Reviews","reviews",31,"2026-09-28T14:31:34.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]