Uber Open-Sources ADR: Enterprise Security for AI Agents at Scale
vijaybolina · x · 2026-08-03
Uber has open-sourced ADR (Agentic Detection and Response), an enterprise security system designed to protect AI agents. Already deployed in Uber's production environment, ADR secures over 50,000 agent sessions per day across thousands of endpoints.
Traditional Endpoint Detection and Response (EDR) tools fall short because they only see surface-level actions like file writes or network calls, missing the underlying agent reasoning and prompt intent. To bridge this gap, ADR captures the full causal chain to provide observability, security benchmarking, and threat detection for enterprise agents like Cursor and Claude Code. The accompanying research paper has been accepted to MLSys 2026.
More from coding & agent
- Awesome AI Memory: A Curated Knowledge Base for LLM & Agent Memory — tom_doerr · 2026-08-03
- Google's Science One Framework: Verifiable Autonomous Research Agents — maier_ak · 2026-08-03
- The Unspoken Rule of AI Coding: Nobody Actually Reviews Generated Code — _jaydeepkarale · 2026-08-03
- The Claude Code Resource Bible: 54 Tools, Agents, and MCP Servers — nikola_mr64990 · 2026-08-03
- Qwen Code v0.21.4 Released: Desktop Web Shell, Subagent Isolation & Cache Optimization — qwen-code-ci-bot · 2026-08-03
- Claude Code Marketplace Launches with 161 Enterprise-Grade Plugins — tom_doerr · 2026-08-03