Max Mathys

I work on AI Control and Monitoring, safeguards, and red-teaming at Lakera AI (acquired by Check Point).

Writing

• Synchronous Control Monitoring: Preventing Harmful Agent Actions in Real Time 2026

Publications

• Breaking Agent Backbones: Evaluating the Security of Backbone LLMs in AI Agents

  ICLR 2026

  Bazinska*, Mathys*, Casucci, Rojas-Carulla, Davies, Souly, Pfister