Uber just opened up early access to ADR (Agentic Detection and Response) — a new system built to bring observability and...

Uber just opened up early access to ADR (Agentic Detection and Response) — a new system built to bring observability and security detection to agentic AI in the enterprise.As AI agents take on more autonomous, tool-using roles, the attack surface changes fast. ADR tackles this with two components:ADR Sensor — captures high-fidelity telemetry from AI agents, including prompts, MCP activity, reasoning traces, tool calls, and execution contextADR Detection — analyzes that telemetry to flag unsafe behavior like credential exposure, prompt injection, data exfiltration, and policy-violating tool useThe code isn't public yet (open-source release is in progress), but the accompanying research paper is already out for anyone who wants a deep dive into the approach.Security for agentic AI is quickly becoming a first-class problem — worth keeping an eye on this one.https://github.com/uber/ADR#AgenticAI #AISecurity #MachineLearning #Cybersecurity #Uber #OpenSource #LLM #AIAgents #EnterpriseAI #Res...

Read Original

Related