Meta plans to extend access to its Llama large language models to more U.S. allies as part of its commitment to providing nations with the tools necessary to protect their national security. France, Japan, and several NATO and European Union institutions will gain access.
“In a world where geopolitical power and national security are deeply intertwined with economic output, innovation and growth, the widespread adoption of open-source models like Llama will be essential to maintaining U.S. and allied AI leadership and ensuring our shared values underpin the systems and standards adopted elsewhere,” Meta said Tuesday.
According to Meta, Llama is designed as an open-source platform that can handle sensitive information without transferring data through third-party artificial intelligence providers. It can also be fine-tuned to meet each government’s specific requirements. The company is collaborating with several organizations to implement Llama-powered solutions for U.S. allies, with support from Accenture, Amazon Web Services and Google Cloud.
Meta’s commitment to ensuring global safety began in 2024, when select U.S. government agencies and contractors, particularly those in defense and national security, were granted access to Llama. The program was later expanded to the Five Eyes nations: Australia, Canada, New Zealand and the United Kingdom.
However, a 2024 Reuters report indicated that an adversary may have already used Llama for military purposes. The report claimed that China used Meta’s model to develop ChatBIT, a military chatbot said to perform at the same level as OpenAI’s GPT-4.