Merlin Research

AI & ML interests

Independent AI safety lab. Stockholm, Sweden. We test deployed LLM agents under adversarial conditions and measure behavioral alignment in production — not in controlled benchmarks.

Team members 2