Adversa AI Trusted AI Blog · November 9, 2026

MCP patch review: bind OAuth issuers and isolate HTTP server instances

Why it matters

Adversa's November roundup collects MCP security research; three maintainer advisories provide immediately actionable checks. The server-everything reference implementation exposed process environment variables over unauthenticated HTTP transports with wildcard CORS; the advisory lists 2.0.1 as the patched release and excludes stdio transport. FastMCP's machine-to-machine OAuth wrappers could send credentials to an authorization server advertised by an untrusted MCP server. Version 4.1.2 adds issuer binding, but operators must configure the expected issuer; an upgrade alone leaves the unsafe compatibility path available. A separate TypeScript server SDK flaw affected shared HTTP transport or server instances in versions 2.0.0 through 2.2.0, allowing response routing across clients with colliding request IDs. Version 2.3.0 rejects that sharing pattern, so affected applications must also construct a server per request or session. These are distinct package and transport conditions, not a single vulnerability affecting every MCP deployment.

My takeaway: Inventory exact package versions and transports. Configure issuer binding, rotate potentially disclosed reusable credentials, replace shared HTTP server instances, and test concurrent clients for isolation. For local HTTP endpoints, verify authentication and origin controls; localhost alone is not an authorization boundary.
Keep exploring

More curated notes connected through Agent Security and AI Engineering.

OWASP GenAI Security Project · guide

OWASP Top 10 for Agentic Applications for 2026

OWASP's community guide organizes agentic-system risk into ten categories, including goal hijacking, tool misuse, identity and privilege abuse, memory poisoning, insecure inter-agent communication, cascading failures, and rogue-agent behavior. It provides a shared taxonomy and mitigation starting point rather than a certification checklist or evidence that a deployed system is secure.

OECD.AI Wonk · guide

A five-step roadmap to closing the AI evaluation gap

The roadmap addresses evaluation results that overstate real-world performance or fail to transfer across deployment contexts. Its five steps balance standardized and local tests, evaluate throughout the lifecycle, build qualified assurance and communication capacity, tailor tests to each value-chain actor and technology, and use a coordinated, trusted process for updating methods.

OWASP GenAI Security Project · tool

FinBot CTF Is Live: A Hands-On Companion to the OWASP GenAI Security Project

OWASP FinBot is a hands-on agentic-security CTF built around a simulated multi-agent financial-services platform with real tool access. Its challenges cover prompt injection, tool misuse, policy bypass, data exfiltration, privilege escalation, remote code execution, shared context, and compromised MCP servers.