Alexander Hagenah / xaitax

AI Security

Offensive security applied to the AI stack: finding real vulnerabilities with frontier models and in-house tooling, working on the security of LLM and agentic systems, and helping boards and governments prioritize where AI risk actually lands.

• AI-assisted discovery of memory-safety and privilege-escalation vulnerabilities (use-after-free, EoP to SYSTEM) and web-class flaws (SSRF, XXE) across leading OS, silicon, and AI-assistant platforms, under coordinated disclosure.
• AI vulnerability-discovery harnesses built for organizational use, run against real targets.
• Work on the security of LLM and agentic systems, including prompt injection, tool misuse, and data exfiltration.
• Contributing co-author on the MIT AI Risk Initiative's 2026 Delphi study Prioritization of Risks from Artificial Intelligence.
• Vetted researcher in Anthropic's Cyber Verification Program; DeepLearning.AI Red Teaming LLM Applications.
• Keynote on Nation-State AI Cyber Offensives at Swiss Cyber Security Days 2025.

Official website: primepage.de