Meta is letting the US military access its AI models for national security purposes
Meta is addressing concerns that its open-source Llama AI models might be helping foreign adversaries by making the Llama series accessible to US government agencies and national security contractors.
In its announcement, Meta revealed partnerships with companies like Amazon, Microsoft, IBM, Lockheed Martin, and Oracle to bring Llama into government use.
Oracle, for instance, is leveraging Llama to analyze aircraft maintenance records, while Scale AI is refining it for targeted support in national security operations. Lockheed Martin, meanwhile, is offering Llama to defense clients for applications such as generating code.
Meta's usual policy forbids developers from using Llama for military, warfare, or espionage-related projects. However, the company is making an exception here, allowing the US military to use Llama for tasks like simplifying complex logistics, monitoring terrorist finances, and bolstering cyber defenses.
A recent report uncovered that Chinese researchers had repurposed Meta's Llama 2 model to develop an AI system for military applications. Meta responded to Reuters, clarifying that this use was "unauthorized" and went against its policy. The company also added that the involvement of an "outdated" version of its open-source model is minor compared to China's extensive AI investment, reportedly in the trillions, aimed at outpacing US advancements.
In its post, Meta emphasized that the US should get ahead in the AI race, highlighting that American open-source models outperforming those from China and other nations would benefit both the US and democracies globally.
Large language models, with their knack for processing massive datasets, reasoning, and generating actionable insights, can undoubtedly bolster aspects of America's and not only national security. But I think the debate over using open AI in defense isn't without reason. Open or "closed," AI for defense stirs up controversy, not just due to its vulnerabilities, like inherent biases and hallucination risks, but also because it leans on personal data.
The US military and defense contractors can leverage Meta's Llama AI
In its announcement, Meta revealed partnerships with companies like Amazon, Microsoft, IBM, Lockheed Martin, and Oracle to bring Llama into government use.
As an American company, and one that owes its success in no small part to the entrepreneurial spirit and democratic values the United States upholds, Meta wants to play its part to support the safety, security and economic prosperity of America – and of its closest allies too.
– Nick Clegg, President of Global Affairs at Meta, November 2024
Meta's usual policy forbids developers from using Llama for military, warfare, or espionage-related projects. However, the company is making an exception here, allowing the US military to use Llama for tasks like simplifying complex logistics, monitoring terrorist finances, and bolstering cyber defenses.
A recent report uncovered that Chinese researchers had repurposed Meta's Llama 2 model to develop an AI system for military applications. Meta responded to Reuters, clarifying that this use was "unauthorized" and went against its policy. The company also added that the involvement of an "outdated" version of its open-source model is minor compared to China's extensive AI investment, reportedly in the trillions, aimed at outpacing US advancements.
Things that are NOT allowed: