NEURALROCKS is an AI security research company.
We study how AI agents fail, and build the infrastructure that records, evaluates, and improves how they behave.
About
NEURALROCKS LLC works in AI/ML security: making the behavior of AI systems measurable, reproducible, and auditable.
Research is the core of the company, and findings are published in the open, with traces and provenance anyone can inspect.
Latest recorded experiment: PPO on MiniGrid-LavaGapS7 under observation corruption →