Is The Pentagon allowed to surveil Americans with AI? (www.technologyreview.com)

The recent conflict between the Pentagon and AI company Anthropic has surfaced critical legal questions regarding the U.S. government's use of AI for surveillance on its citizens. The Pentagon sought to leverage Anthropic's AI, Claude, to analyze bulk commercial data on Americans, which prompted Anthropic to refuse, insisting its technology not be employed for domestic surveillance or autonomous weapons. Following failed negotiations, the Pentagon labeled Anthropic a supply chain risk, a designation typically aimed at foreign threats. In contrast, OpenAI, which signed a controversial contract allowing the Pentagon to utilize its AI for "all lawful purposes," faced significant public backlash, leading to a wave of subscription cancellations. OpenAI later revised its contract to explicitly prohibit its AI from being used for domestic surveillance. This debate underscores the significant implications for privacy and national security as current laws have not adapted to the rapid advancements in AI capabilities. Legal experts point out that much data accessible to the government is technically not regulated, raising concerns about unwarranted surveillance practices. While OpenAI emphasizes new safeguards and the intention to prevent surveillance misuse, ambiguity persists around the contract's language, which may still allow for surveillance under the guise of lawful purposes. As legislators like Senator Ron Wyden push for clearer guidelines on mass surveillance, the broader AI community watches closely, understanding that this situation could redefine the intersection of technology and privacy rights.