OpenAI moved to renegotiate a recently announced Pentagon agreement after internal and public criticism, while the Defense Department simultaneously canceled fellowship programs at a group of elite U.S. colleges. OpenAI’s memo said it would add explicit prohibitions on domestic surveillance in revised contract language; the Pentagon has meanwhile labeled companies as supply‑chain risks in a standoff that began with Anthropic’s refusal to accept certain military usage terms. The dispute has immediate consequences for university research and training pipelines: the Pentagon’s cancellation of fellowships at 13 institutions severs a funding and talent channel and signals increased vetting of academic partners. Universities face a hard choice between defense research income and institutional commitments on civil liberties and research ethics. Legal and policy experts warned the episode raises governance questions about who controls dual‑use AI applications, the enforceability of contractual “red lines,” and the risk that defense procurement decisions will reshape campus research priorities.