Days after a Pentagon contract that made global headlines, OpenAI is in early-stage discussions with NATO about a potential deployment on unclassified alliance networks. The Wall Street Journal and Reuters reported the discussions Monday, March 3.
Scope matters here. The talks are limited to unclassified networks, which represents a significant constraint on what any deployment would involve. CEO Sam Altman had previously suggested broader classified access; according to the Wall Street Journal, which cited people familiar with an internal company meeting, Altman walked back that characterization. NATO has not confirmed the reported discussions.
The timing tracks with a broader pattern. The Anthropic supply chain designation is generating vendor switches inside US defense firms. That same dynamic is clearing competitive space for OpenAI in government contracting. Whether those two things are related is a question the evidence doesn’t answer yet. What the evidence does show is that OpenAI’s government footprint expanded in at least two directions this week — deeper into US defense infrastructure and wider across allied institutions.
Early-stage discussions with NATO are not a contract. The distance between exploratory conversations and a deployed capability is significant, particularly given the alliance’s member-state governance requirements and procurement timelines. Anyone extrapolating a near-term NATO deployment from Monday’s reporting is moving well ahead of what’s confirmed.
That said: the direction is clear. Five days into this week, OpenAI’s government AI position looks materially different than it did last Friday.