What happened
On about 24-30 September 2026 Rep. August Pfluger (R-TX) introduced H.R.10594, the Reliable Artificial Intelligence Research Act of 2026, directing the Department of Homeland Security to run at least two prize competitions: one on AI interpretability (understanding how AI systems make decisions) and one on adversarial robustness (resisting manipulation/attacks), in high-impact high-risk use cases, with $10M authorised for FY2027-2031. DHS must consult NIST, Commerce, the National Cyber Director and NSF; red-teaming phases are contemplated.
Why it matters
It is an early-stage legislative effort to build federal capability in AI transparency and robustness via public competitions — relevant to how US government AI-safety research funding and procurement direction are forming, though it creates no direct compliance burden.
Action needed
AI researchers, red-team companies and defence-tech vendors should monitor the bill; if enacted, NIST/DHS-linked interpretability and robustness prize competitions could influence federal procurement expectations for AI trustworthiness.