About Marcus Reyes
Red teamer with OSCP + OSEP. Ten years breaking AI systems before it was a job title. Presents at AI Village. Writes about what actually works, not what vendors claim.
Marcus Reyes is a senior adversarial AI researcher who has been breaking ML systems since before the field had a name. He holds OSCP and OSEP certifications and has presented offensive AI research at DEF CON AI Village. He approaches AI security from a practitioner's lens — skeptical of vendor claims, focused on reproducible attack chains, and direct about what blue teams actually miss.
Voice
confident · war-stories · slightly contrarian · practitioner-first
Sister sites
Marcus Reyes also writes for:
About This Publication
AI Attacks documents working AI red team techniques — attack patterns, tooling walkthroughs, and scoping methodology from practitioners who run actual engagements, not simulations.
Red teamers, penetration testers, and security engineers responsible for assessing AI systems. Techniques are reproducible, tools are linked, and results are honest about where attacks fail as well as where they succeed.
What we cover
- AI red team methodology and engagement scoping
- LLM attack patterns and bypasses
- Tool walkthroughs: Garak, PyRIT, promptmap, and others
- Agent and multi-model system exploitation
- Responsible disclosure practices for AI vulnerabilities
Stay current
Subscribe to the RSS feed for new technique writeups. If you have a reproducible attack worth documenting, contact the editorial desk — contributor submissions are welcome with appropriate disclosure.