Tracking AI governance, autonomous weapons systems, frontier model development, compute policy, and the integration of machine intelligence into strategic competition.
Bipartisan legislation would require frontier AI labs to submit capability evaluations to a federal AI Safety Board before deploying models above a compute threshold, drawing both support and criticism.
Project Maven, the DoD's flagship AI targeting initiative, is being extended to ground-based and maritime tactical systems following successful integration with drone swarm coordination in classified exercises.
New open-source analysis of PLA technical papers reveals significant investment in AI-enabled command and control systems, including automated threat classification and response recommendation engines.
The Act's broad carve-out for national security applications leaves member states free to deploy AI systems in defense contexts without the oversight framework applied to civilian uses, raising interoperability concerns.
A recently declassified joint assessment found that large language models used in OSINT synthesis pipelines produced plausible but factually incorrect assessments in 12% of test cases without adequate human review.
The race to 10^27 FLOP training runs is creating governance gaps as compute capacity diffuses globally. Experts warn that export controls alone cannot contain frontier capability development past 2027.