Pentagon: OpenAI and Google's AI to Secure Its Networks

The U.S. Cyber Command is ramping up its efforts to integrate advanced AI models into the Pentagon's and NSA's top-secret networks. This initiative aims to leverage the capabilities of AI models developed by tech giants like OpenAI and Google to enhance the security of sensitive digital systems.

In light of the rapid evolution of AI technologies capable of detecting security vulnerabilities more effectively than human hackers, Cyber Command views these tools as crucial for improving threat detection and accelerating decision-making in cyber operations.

A new task force has been established to facilitate the deployment of these AI models on highly classified networks. This decision, reported by Politico, is based on statements from three sources close to the project. General Joshua Rudd, who leads both the NSA and Cyber Command, recently announced the creation of this task force in an internal email.

The goal is to assess how AI models from Silicon Valley companies can be securely integrated into "high-side" systems, which are those with the highest classification levels. The NSA's AI Security Center will provide the necessary technical expertise, while an officer from Cyber Command will lead the effort.

The development of AI models capable of detecting vulnerabilities faster than the best human hackers has been a triggering factor. In April, Anthropic launched its Claude Mythos model but limited access due to potential risks to the economy and national security. OpenAI has also announced models with similar capabilities, and Anthropic anticipates that comparable tools could be widely available within the next six to twenty-four months.

Lieutenant General Charles Moore, former deputy commander of Cyber Command, emphasized the importance of this task force, stating that AI tools are now essential for threat detection, vulnerability prioritization, and rapid decision-making in cyber operations, both offensive and defensive.