MOVE78RESEARCH ENGAGEMENTS / COLLABORATIONS
Bring us a difficult question.
Let’s build the evidence to answer it.
We work with researchers, developers, and safety organizations to understand increasingly capable AI systems. Engagements connect rigorous evaluation with hands-on adversarial expertise.
A shared question. A rigorous investigation.
Joint studies, evaluation methods, and original research on agent behavior and embodied intelligence. We are interested in work that makes uncertain capabilities and failure modes observable.
Find out what the system can really do.
Task-driven evaluations of autonomous cyber capabilities, operational judgment, and embodied reasoning. We help distinguish promising demonstrations from dependable behavior.
Turn a failure mode into an engineering question.
Adversarial investigation of the trust boundaries around intelligent systems, with a focus on practical controls and evidence that those controls hold.
HOW WE WORK
Agree on the system, threat model, scope, and evidence that would answer the question.
Design controlled experiments and inspect behavior with an operator’s attention to detail.
Document observations, limitations, and implications for research and engineering.
THE NEXT QUESTION