bits to atoms: red-teaming the robot age

Field notes

The lab organizes its proposed work into three stages: threat modelling, red-teaming, and technical mitigation. Its risk map is intended to score physical-AI hazards by impact, tractability, and neglectedness, while experiments will use scenarios drawn from that map rather than convenience cases. The site says threat modelling is under way and reports secured grants, a research roadmap, and fundraising for experimental runs, but does not yet document completed tests or deployed safeguards.