Pentagon AI Deals: Data Loophole & Privacy Risks

The Pentagon has signed deals with eight AI companies, including SpaceX, OpenAI, and Google, to use their advanced AI on classified networks. This follows a dispute with another AI company, Anthropic. Anthropic's large language model, Claude, was part of a pilot program with usage restrictions. These banned its use for mass domestic surveillance and fully autonomous weapons. The Pentagon wanted to remove these restrictions. When Anthropic refused, the Pentagon moved to blacklist the company, and Anthropic has since sued. The Pentagon's Chief Technology Officer says existing laws already bar mass surveillance of Americans. However, this relies on the idea that commercially available data, like communications metadata purchased from brokers, doesn't count as surveillance. Foreign intelligence programs also acquire Americans' information without warrants. Large language models increase the risk to privacy by making it easier to analyze vast amounts of data and infer details about people. This matters because the misuse of such capabilities poses serious risks to privacy and civil liberties.