Claude projects silently enable account memory sharing, breaking promised isolation
bcRIPster · reddit · 2026-09-28
A Redditor discovered Anthropic quietly added a "Use account memory" toggle inside every Claude project, default-on for existing projects — bridging project memory with account-level memory. They found real cross-contamination (fiction character names mixed with job-search and research material) and spent an hour cleaning up. Anthropic's help docs still claim project memory is isolated, and the toggle apparently shipped with the Sept 16-17 platform unification without any announcement. The author urges lawyers, consultants, and researchers using projects for isolation to check settings and turn it off, and has filed formal feedback.
More from Models
- TeleOCR Trends on Hugging Face: A Qwen2.5-VL-Based Chinese Document OCR Model — XingChen-AGI · 2026-09-28
- Kaggle Game Arena: Evaluating LLMs via Head-to-Head Chess, Poker, and Werewolf — kaggle · 2026-09-28
- Perplexity CEO: still using sol 6 for knowledge work — cheap, fast, great compaction — gabriel1 · 2026-09-28
- NerfBench's First Results Find No Nerf: Claude Opus 5.5 Dips Just 0.8% vs Launch — alejandroll10 · 2026-09-28
- Most humans can read this image instantly — most AI vision models can't — JeremyNguyenPhD · 2026-09-28
- AI-generated 7-minute SQLite repo explainer stuns with coherent code walkthrough — deedydas · 2026-09-28