AI Doomsday Is Not Inevitable, but Human Accountability Is

Summary

The Intercept Briefing examines how the U.S. government increasingly relies on private companies for surveillance, immigration enforcement, and military operations. Speakers Sam Biddle and Jason Koebler describe a market in which local police, federal agencies, and contractors can buy access to license-plate data, cellphone locations, browsing histories, financial records, facial-recognition systems, and other information. They argue that outsourcing can bypass warrant requirements and weaken public oversight, while data collected by local departments can flow upward to federal agencies. The discussion highlights Flock cameras, Palantir, Mobile Fortify, predictive targeting teams, and lesser-known contractors involved in locating and identifying immigrants. It also addresses the military use of AI systems. The speakers say Anthropic, OpenAI, Google, and xAI signed Pentagon contracts that went beyond selling software and included briefings, training, threat assessments, and advice on military use, although the contracts were heavily redacted. They say Claude has been linked to targeting and planning in the context of U.S. operations involving Iran, but the precise level of autonomy and the circumstances of specific strikes remain unclear. The discussion contrasts two risks: companies may be directly involved in warfare, or they may lose visibility after handing systems to government customers. Both raise accountability concerns. Possible responses mentioned include independent audits, limits on frontier AI development, local and state privacy rules, public opposition to surveillance infrastructure, and opting out of airport biometric identification. The speakers reject the idea that AI itself wants to destroy humanity, arguing that focusing on machine motives can obscure the people and institutions that build, buy, deploy, and regulate these systems. They conclude that current harms, including surveillance and AI-assisted violence, deserve attention alongside hypothetical existential scenarios, and that cross-partisan resistance to tools such as Flock could support broader privacy protections.