Meta staffers internally warned Muse AI agent was accessing encrypted private messages
Ars Technica AI · rss · 2026-11-19
After early users of Meta's Muse AI agent complained it was spying on private messages and behaving unpredictably, Reuters obtained internal chats showing Meta staff raised the same alarms. During testing, staffers warned Muse appeared to access encrypted private messages users never expected it to review, possibly mixed up conversations between users, and risked exposing sensitive information — with some outputs responding to questions nobody had asked.
More from Companies & People
- Lovable CEO Anton Osika on turning two: don't say "someone should," say "let's do this" — mmurph · 2026-11-19
- Stanford AI100 Releases Third Report on the Scaling Era and Rise of Generative AI — StanfordHAI · 2026-11-19
- Travis Kalanick calls Jensen Huang's ouster "an op," warns "don't let Nvidia go to the suits" — zealcaiden · 2026-11-19
- Ex-Nvidia employee says he was fired for advising board to gather facts first — zealcaiden · 2026-11-19
- Burkov begins 'The New Hundred-Page Machine Learning Book' after 7 years — dioscuri · 2026-11-19
- vLLM team's Inferact to build an autonomous inference system around vLLM — vllm_project · 2026-11-19