AI Firm Anthropic Monitors Activists, Reports to Police

Investigative reporting has detailed security officials at Anthropic, the creator of Claude AI, engaging in monitoring and surveillance activities, including reporting AI activists to the police for potential future crimes. This comes amid Anthropic's dispute with the Pentagon over military AI deployments.

Anthropic asserts that these actions are part of a physical-security function to ensure employee safety, comparable to any large company with a global workforce, and disputes claims of building a predictive surveillance system for activists. The company's security team tracks individuals who oppose rapid AI development near its physical locations.

Recent reports indicate Anthropic has made routine reports to police about threats against employees, seeking to identify escalation patterns early. A job posting for an 'enterprise intelligence specialist' described a role involving the identification and investigation of global threats, including activism, which Anthropic stated was a general threat category monitored for employee safety.

The trend of predictive policing, or 'precrime,' is growing, with AI tools analyzing data to forecast criminal activity. Anthropic's Claude AI and other intelligence tools reportedly allow it to share conversation data with police without court orders, raising concerns about due process and withholding evidence.

Concerns are also raised about how much access Anthropic's AI has to user data, including social media and potentially linked email accounts, and whether this information could be used to track activists. The company states its monitoring of Claude aims to detect malicious activity like influence operations and scams, fulfilling an obligation to disclose risks and make AI models safer, but the line between malicious activity and activism remains a question.