BerriAI / litellm
github.com·3d ago
Developer Tools
GitHub Trending·Aug 4·5 min read
uber / ADR
TL;DR
ADR (Agentic AI Detection and Response) is a security system developed by Uber to protect AI agents used in enterprises. It enhances security through observability, threat detection, and prevention of unsafe actions.
✦ Why It Matters
Evaluate your AI agents' security using ADR to ensure safe operations in your organization.
Key Takeaways
- Utilize ADR to monitor AI agent activities effectively.
- Test agent security with ADR-Bench's extensive task coverage.
- Leverage the dual-agent detection architecture for improved threat identification.
- Stay updated on the release of the ADR prevention component.
How It Works
ADR operates by integrating four key capabilities: observability captures and normalizes telemetry from various AI tools, benchmarking tests agent security under realistic conditions, detection identifies risky behaviors through a two-tier architecture, and prevention aims to stop unsafe actions before they occur.