Meta Platforms and Scale AI have joined forces to develop Defense Llama, a military-focused adaptation of Meta's open-source Llama large language model designed to support American national security operations.
The specialized AI system, available exclusively in controlled U.S. government environments through Scale's Donovan platform, enables military planners and intelligence analysts to apply generative AI to defense-specific tasks such as planning operations and analyzing adversary vulnerabilities.
According to Scale AI, Defense Llama was trained on a comprehensive dataset including military doctrine, international humanitarian law, and Department of Defense guidelines to ensure its outputs align with ethical principles while maintaining accuracy for defense applications. The model has undergone evaluation using DoD-specific benchmarks to verify its grasp of military terminology and protocols.
"Our open source Llama models have tremendous potential to support America's safety and security and keep our country globally competitive," said David Wehner, Meta's Chief Strategy Officer, discussing the partnership with Scale AI.
This collaboration emerges just months after Meta changed its policy in November 2024 to allow U.S. government agencies and defense contractors to use its AI models, opening the door for its technology to play a key role in military and national security efforts. The company previously prohibited military uses in its terms of service.
The Defense Llama partnership coincides with Meta's recent collaboration with defense contractor Anduril Industries to develop military products, including the EagleEye system—an AI-powered helmet with virtual and augmented reality features that enhances soldiers' vision and hearing on the battlefield. The helmet integrates Meta's Llama AI model and draws on technology from Meta's Reality Labs.
These developments reflect a broader trend of defense-tech companies leveraging classified military data to train AI models, with the defense industry's substantial budgets attracting mainstream AI companies. OpenAI similarly announced a partnership with Anduril in December 2024 focused on developing AI systems to defend against drone attacks, marking a significant shift from its previous policy of not working with the military.