Google DeepMind published a technical roadmap aimed at protecting against “rogue” AI agents, shifting from a narrow focus on alignment to a layered insider-threat style security model. The 35-page report outlines procedures to monitor agent behavior and limit access to tools and data. At the federal level, a separate account from the Pentagon said AI use has surged but adoption still lags, with Chief Technology Officer Emil Michael reporting millions of personnel potentially using commercial AI tools—yet still under half of agency employees. The Pentagon highlighted partnerships and new operational integrations, while noting lingering uncertainty about work product quality. For higher education, the parallel message is direct: AI adoption is accelerating faster than governance, controls, and auditability—raising the stakes for cybersecurity, privacy, and responsible use on campus.