Syed Anas Mohiuddin

• Security of AI agent systems and the protocols they use, mainly MCP and A2A
• Coordinated vulnerability disclosure in MCP servers
• How vulnerability records lose reporter credit as they move between databases

Based in Chicago, Illinois, United States. I run Cognivators, which sells Agent Surface and security reviews for teams that ship MCP servers and AI agents.

1. CVE-2026-14540

   Google: Server-side request forgery via unrestricted HTTP redirection in MCP Toolbox. Rated 8.0 High (CVSS 4.0, assigned by Google).

   Read the write-up

2. CVE-2026-97228

   Rapid7: GraphQL query injection in the export status lookup of Rapid7 Bulk Export MCP. Rated 2.7 Low (CVSS 3.1).

   Read the write-up

3. github-mcp-server #3056

   Patch authored by me and merged by GitHub on 2026-08-18. A hardening change, with no CVE.

   Read the write-up

1. 2025-09

   Founds Cognivators LLC in Chicago.

2. 2026-05-06

   First public release of mcp-safeguard on PyPI.

3. 2026-05-24

   Publishes the Protocol Pivoting preprint.

4. 2026-06-02

   Files IETF Internet-Draft on MCP security considerations.

5. 2026-06-18

   Google merges the fix for the SSRF he reported in MCP Toolbox.

6. 2026-07-31

   CVE-2026-14540 published by Google, naming him as finder.

7. 2026-08-18

   GitHub merges his patch to github-mcp-server (PR #3056).

8. 2026-08-26

   Publishes Never Emitted, on lost reporter credit in vulnerability records.

9. 2026-09-01

   Advisory GHSA-q3gg-rgxh-gh64 published for 1Password-MCP, finder credited.

10. 2026-09-25

    CVE-2026-97228 published by Rapid7, naming him as finder.

11. 2026-10-05

    Ars Technica covers the research.

12. 2026-10-23

    Speaks at AGNTCon / MCPCon, San Jose.