https://aegisgatesecurity.io/tags/owasp/Recent content in Owasp on AegisGate — Secure Every AI InteractionHugo -- gohugo.ioen-usWed, 14 Oct 2026 00:00:00 +0000https://aegisgatesecurity.io/blog/mcp-tool-poisoning-the-attack-hidden-in-plain-text/Wed, 14 Oct 2026 00:00:00 +0000https://aegisgatesecurity.io/blog/mcp-tool-poisoning-the-attack-hidden-in-plain-text/In April 2025, Invariant Labs demonstrated something unsettling: you don’t need to hack an AI agent to make it misbehave. You just need to describe a tool. They called it tool poisoning — hidden instructions embedded in MCP tool descriptions that manipulate agent behavior. No authentication bypass. No exploit. Just text in a JSON field that the agent reads as context. The MCP spec assumes tool metadata is trustworthy. It isn’t.