Early‑stage AI agents that autonomously log into learning platforms and complete coursework are emerging and raising immediate academic‑integrity concerns. Products like a reportedly agentic tool nicknamed “Einstein” claim to access Canvas, watch lectures, write papers and submit assignments without human input. At the same time, users experimenting with always‑on agents report fragility, unpredictable behavior and the need for intensive oversight. Educators and IT leaders face a fast‑moving detection and policy challenge: detection alone may be insufficient and analog assessment is only a partial defense. Institutions must evaluate the capabilities of agentic tools, reassess assessment design, and accelerate policy and technical responses that balance academic standards with pedagogical priorities. Campus leaders should brief faculty on evolving threats, pilot detection and honor‑code interventions, and coordinate with legal and privacy offices on data‑use and vendor risk.