Brooks McMillin
AI Security Researcher & Infrastructure Security Engineer
I lead infrastructure security at Dropbox, focusing on AI agent security, LLM development tooling, and securing production AI systems. My team builds frameworks that help engineers safely ship AI-powered features while defending against emerging threats to autonomous systems and ML infrastructure.
A Beginner's Guide to Safe LLM-Assisted Development
A complete beginner's guide to setting up every safety layer from the Coding Safer with LLMs post: pre-commit hooks, local review agents, CI workflows, and CLAUDE.md — starting from scratch.
Read ArticleCurrent Focus
AI Agent & Infrastructure Security
Building security frameworks for AI agents, developing LLM security tooling for engineering teams, and securing the infrastructure that powers production AI systems at scale.
- Designing security architectures for autonomous AI agents — sandboxing, permissions, and runtime controls
- Building LLM security tooling that integrates into developer workflows for safe AI feature development
- Threat modeling for MCP, multi-agent systems, and tool-use patterns in production environments
- Securing infrastructure for AI/ML platforms — identity, access control, and data protection
Featured Projects
Agent Framework
productionA production-ready framework for building LLM agents with the Model Context Protocol (MCP), featuring OAuth 2.0 with PKCE, persistent memory, and extensible tool architecture.
- Production-ready framework for building LLM agents with MCP
- Full OAuth 2.0 with PKCE and dynamic client registration
Secure Task Management Suite
productionA comprehensive secure task management platform with OAuth 2.0 authorization, Python SDK, and MCP server integration for LLM frontends.
- Full OAuth 2.0 authorization server with PKCE support
- Comprehensive security testing suite with Vitest
Secure Communications Suite
productionA dual-language secure communications suite providing SMS capabilities through GSM modems, built for security operations and incident response scenarios.
- Cross-platform GSM modem interface (Go + Python)
- Real-time SMS communication with security considerations
ReMarkable Research Security Toolkit
productionA comprehensive toolkit for secure research paper management on reMarkable tablets, featuring AI-powered classification and automated organization.
- Automated research paper classification using AI
- Secure PDF processing pipeline with content validation
Recent Appearances
Real patterns and failures from building specialized agents with shared infrastructure, covering capability bounding, prompt injection detection, memory isolation, and OAuth device flow
LLMs Will Never Be Fully Secure
podcastDiscussion on malicious MCP servers, recurring security mistakes in AI tooling, prompt injection persistence, and practical strategies for safe AI deployment
Security audit of MCP servers and their OAuth implementations, demonstrating that 90% of vulnerabilities reflect longstanding security principles amplified by AI agents
A three-layer defensive framework for catching security mistakes introduced by LLM-assisted code generation before they reach production
Let's Connect
Interested in AI security research, collaboration opportunities, or speaking engagements? I'd love to hear from you.
More About Me