A newly announced agreement between OpenAI and the US Department of Defense underscores a growing divide in the AI sector—not over whether military cooperation should exist, but over who ultimately governs the limits of how advanced AI systems are used in classified environments.