They gave 4.1 to U.S. State Department

OpenAI has initiated a landmark pilot program with the U.S. Department of State, granting the diplomatic agency exclusive early access to GPT-4.1, a yet-to-be-publicly-released iteration of its flagship model. The arrangement, reported as part of a strategic partnership, allows State Department analysts to utilize the advanced AI for parsing dense international treaties, summarizing lengthy negotiation transcripts, and identifying key points across multiple languages. This move signals OpenAI's deepening engagement with governmental entities and a strategic pivot toward applying its most capable models to national security and foreign policy challenges, areas traditionally guarded against commercial AI involvement.

The GPT-4.1 model provided is understood to offer enhancements in reasoning over long, complex documents and improved accuracy in nuanced language interpretation—critical for diplomatic work. The pilot focuses on back-office analysis rather than live negotiation support, aiming to augment human expertise by quickly surfacing patterns and inconsistencies across thousands of pages of diplomatic text. This collaboration sets a precedent for how frontier AI labs might work directly with governments, raising questions about model access tiers, the control of advanced capabilities, and the potential for an 'AI divide' in statecraft. The outcomes of this pilot could influence how other agencies and allied governments seek to integrate similar AI tools into their own strategic processes.