Meta to make Llama models available for US national security applications

To address concerns about the use of its “open” AI by foreign adversaries, Meta has announced that it will be providing its Llama series of AI models to U.S. government agencies and contractors involved in national security efforts. This move is aimed at bolstering national security and defense applications within the U.S.

Partnerships with companies such as Accenture, Amazon Web Services, and Lockheed Martin have been established to facilitate the integration of Llama into government agencies. For instance, Oracle is utilizing Llama for processing aircraft maintenance documents, while Scale AI is tailoring Llama to support specific national security missions.

Despite Meta’s usual policy prohibiting the use of Llama for military, warfare, or espionage projects, exceptions have been made for government agencies and contractors in countries like the U.K., Canada, Australia, and New Zealand. This decision comes in the wake of reports that Chinese researchers associated with the PLA had utilized an outdated Llama model for defense-related purposes, sparking debates on the risks and benefits of open AI.

A recent study by the AI Now Institute highlighted the potential dangers of utilizing AI for military intelligence, surveillance, and reconnaissance due to vulnerabilities like biases and data exfiltration. The debate over the use of AI in defense applications has led to protests by employees at major tech companies like Google and Microsoft against their involvement in creating AI tools for the military.

Meta argues that open AI can enhance defense research and advance America’s economic and security interests. However, the adoption of AI technology by the U.S. military has been slow, with skepticism surrounding its return on investment. Despite some branches of the U.S. armed forces, like the U.S. Army, deploying generative AI, broader acceptance of AI in defense remains a subject of ongoing discussion and scrutiny.