The Register analyzes leaked Claude Code source code, revealing extensive data collection capabilities: persistent telemetry sent to Anthropic on every launch (user ID, email, session ID, feature flags), full retention of every file Claude reads, a hidden "KAIROS" headless daemon, an unreleased "autoDream" background agent that scans all session transcripts, and remotely-managed settings that can push .env variable changes without user notification. An "undercover.ts" file instructs Claude to hide AI authorship in open-source commits. The article also covers Anthropic's ongoing lawsuit against the US Department of War over AI service bans in classified environments.
Safety
Claude Code source leak reveals how much info Anthropic can hoover up about you and your system
Leaked Claude Code source reveals Anthropic collects persistent telemetry on every launch and runs undisclosed background daemons (KAIROS, autoDream) that scan transcripts and enable remote settings manipulation.
Friday, April 3, 2026 12:00 PM UTC2 MIN READSOURCE: The RegisterBY sys://pipeline
Tags
safety
/// RELATED
Infrastructure4d ago
micro-benchmarks don’t tell the whole story
Rust's HTTP frameworks dominate raw throughput benchmarks (316k req/s), but the analysis reveals micro-benchmarks measure only socket performance and ignore real-world application bottlenecks that determine actual user experience.
ModelsApr 28
BiTA: Bidirectional Gated Recurrent Unit-Transformer Aggregator in a Temporal Graph Network Framework for Alert Prediction in Computer Networks
BiTA, a hybrid GRU-Transformer architecture, improves network attack detection by capturing temporal dependencies in alert patterns—outperforming existing temporal graph models.