IN Brief:
- Eight AI companies will deploy capabilities on US classified network environments.
- The agreements cover IL6 and IL7 systems for operational, intelligence, and enterprise use.
- The move creates new production pressure around secure AI deployment, accreditation, and vendor resilience.
The Pentagon has entered agreements with eight AI companies to deploy advanced capabilities on classified network environments, moving commercial artificial intelligence deeper into defence infrastructure.
SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, Amazon Web Services, and Oracle are included in the agreements. Their systems will support Impact Level 6 and Impact Level 7 environments, placing AI tools closer to operational, intelligence, and enterprise workflows.
The agreements build on GenAI.mil, the Pentagon’s official AI platform, which has already reached more than 1.3 million personnel. The platform has generated tens of millions of prompts and deployed hundreds of thousands of agents in its first five months of use.
The inclusion of multiple vendors also reduces dependence on a single supplier. The Pentagon is building a more diverse AI stack across models, cloud infrastructure, compute, and software environments, with classified deployment becoming a central test of supplier maturity.
Classified software as a production system
AI in classified environments requires more than model access. The production system includes secure hosting, data controls, identity management, audit trails, model evaluation, cyber monitoring, and update processes that can operate inside government networks.
That creates a different industrial rhythm from traditional platform production. Software suppliers have to maintain controlled pipelines, patch security issues, document model behaviour, and support continuous assurance. Capability will evolve after deployment, so accreditation processes need to support change without allowing uncontrolled risk.
Compute and infrastructure providers also move closer to the mission layer. Chips, cloud services, classified network architecture, and model-serving environments all become part of the operational product.
The agreements put AI suppliers into a defence market shaped by security, continuity, and governance as much as performance. The companies that can deliver reliable capability under classified constraints will have an advantage as military AI moves from experimentation into routine operational use.
