OpenAI hires ex-defense secretary Mark Esper as its military AI work turns into a core business

OpenAI has hired Mark Esper, the former US Secretary of Defense, as a strategic advisor for its military AI initiatives, according to Crypto Briefing, which reports the appointment was announced on November 2, 2026. The move comes a few years after OpenAI's usage policies banned military and warfare applications outright. Those prohibitions were removed in 2024.

The pivot has moved quickly since then. On February 9, 2026, OpenAI announced that a custom version of ChatGPT would be available on GenAI.mil, the US military's generative AI platform, for up to 3 million users. On February 27, it signed an agreement with the Department of Defense to deploy its models in classified environments. According to the report, that agreement sets limits against mass domestic surveillance and against autonomous weapons. OpenAI has also gathered unclassified defense contracts worth up to $200 million.

Our thesis is that hiring a former Pentagon chief is a statement of intent. Advisors with that profile exist to open doors in procurement and policy, so OpenAI now treats defense as a business line, not a side project. Crypto Briefing points to the $200 million ceiling as the number to watch. If the classified deal and Esper's network turn into larger awards, defense revenue could grow well beyond it.

The short-term concern is about governance, not about capability. The source makes a fair point. Limits written into an agreement are only as durable as the enforcement and transparency behind them. OpenAI already rewrote its rules once, when it dropped its 2024 ban. Its red lines on surveillance and autonomous weapons will face pressure as the government work deepens, and right now there is no public mechanism that lets outsiders check whether they are respected.

Our reading: this fits the pattern we have tracked, from export controls on Anthropic's Mythos to the CNAS report on China. Frontier models are dual-use assets, and states will want them. Democracies using the best AI for logistics, analysis and defense is not inherently bad, and a US lab staying out of the field would not stop rivals from entering it. What matters is who governs the technology and how visibly. Our long-term optimism depends on AI's benefits arriving without a normalized surveillance state or delegated lethal decisions along the way. OpenAI should make its limits auditable before the contracts get bigger.