The Coalition for Secure AI (CoSAI) has released a comprehensive whitepaper addressing Model Context Protocol (MCP)—the emerging standard that's rapidly becoming the backbone of AI agent infrastructure.
CoSAI Project Governing Board (PGB) co-chair Omar Santos explains how to operationalize the CoSAI Risk Map (CoSAI-RM) through a structured, persona-based approach that moves beyond ad-hoc security measures. Santos shares a practical, four-phase approach for turning CoSAI-RM from a conceptual model into an actionable risk management strategy, from persona-based risk […]
The Coalition for Secure AI (CoSAI) has released the AI Incident Response Framework, Version 1.0, which provides security teams with the tools, knowledge, and structured approaches they need to protect AI systems from emerging threats.
This strategic update was written by a group of CoSAI Project Governing Board members with input from the CoSAI Technical Steering Committee and workstreams. It previews upcoming papers. This update does not represent the views of all members of the CoSAI project.
Blog Post: CoSAI Principles for Secure-by-Design Agentic Systems - practical implementation strategies aimed at agent developers, adopters, and security engineers.
Researchers at HiddenLayer, a CoSAI sponsor, have developed the first universal and transferable prompt injection technique that bypasses instruction guardrails across major AI models.