AdversarialMinds · Steve Brodson

AI needs security. I build both.

I'm a cybersecurity architect focused on AI safety and security. I experiment with AI systems, consult with organizations navigating AI risk, and teach practitioners how to think clearly about threats that don't fit traditional security frameworks.

Read the writing → Subscribe
01 — Why it matters

It comes down to trust.

The box gets checked, the tool gets bought, but the real problems go unaddressed.

The organizations that navigate this well won't be the ones with the most sophisticated AI, or the fanciest tools. They'll be the ones with the most honest internal view of what they've deployed, what it can actually do, and what they're accepting when they run it.

That's the work. Everything else is surface area.

— Steve
02 — Latest writing

Field notes from the practical side of AI security.

2026 · 04 · 13
The AI Safety Conversation We Need to Have
The AI Safety Conversation We Need to Have
The AI safety debate is stuck on science fiction while real problems compound in production. Shadow AI, unsolved vulnerabilities, and the multiplication of risk in agentic systems demand a practical conversation.
→
2026 · 04 · 16
AI Security - Five questions that don't have answers
AI Security - Five questions that don't have answers
AI systems fail in ways traditional security wasn't built to handle. Five open problems in AI security — what they are, why the old playbook doesn't fit, and why solving them may require thinking differently.
→
2026 · 04 · 16
Traditional Cybersecurity Isn't Enough for AI
Traditional Cybersecurity Isn't Enough for AI
Traditional cybersecurity frameworks were built for deterministic systems. AI introduces non-determinism, and that changes how you need to think about threat modeling, testing, and defense.
→
2026 · 04 · 14
Phishing Defenses Are Falling Behind AI
Phishing Defenses Are Falling Behind AI
Phishing has been the leading cause of breaches for decades. Better filters, authentication, and training have helped and still do. AI is making that work exponentially harder.
→
2026 · 04 · 12
The Axios Compromise Is the Warning Shot. AI Is the Real Story.
The Axios Compromise Is the Warning Shot. AI Is the Real Story.
The Axios npm compromise on March 31 hit a library with 100 million weekly downloads. The attack wasn't technically sophisticated — it exploited assumptions baked into how we build software. AI is about to change what both attackers and defenders can do with those assumptions.
→
All posts →
03 — What I do

Three ways I help organizations work through this.

01 · Consulting

Working alongside teams.

Embedded work with security, engineering, and leadership — mapping shadow AI, threat-modeling the non-deterministic surface, and setting governance that survives contact with production.

02 · Speaking

Plain-English keynotes.

Talks for boards, industry summits, community groups, and internal all-hands. No jargon, no doom — a clear picture of what's actually happening and what matters to decide.

03 · Tooling

Building what's missing.

Small, practical tools that help teams see and govern the AI already running in their environment. Released quietly, maintained honestly.

04 — Newsletter

Bi-weekly analysis, in your inbox.

Essays on what's actually happening in AI security, updates on the latest exploits, and briefings on new technologies — what the tools catch, what they don't, and how to make cleaner decisions inside genuine uncertainty.

Subscribe →