DeepYard
P

PIPES

Secure agent perception through data provenance tracking and authority controls

Open SourceFree

About

PIPES (Provenance and Priors for Agent Perception Security) is a research framework that protects AI agents from state-corruption attacks by tracking data provenance and enforcing information authority controls. It ensures agents can verify the trustworthiness of their inputs before making decisions, addressing a critical security gap in autonomous systems that rely on external data sources.

Details

Type
Integrations
Language

Tags

autonomousobservabilityframeworkopen-sourceresearch