EFF Warns Congress: AI Surveillance Threatens Constitutional Rights

Government AI deployment needs constitutional guardrails

The Electronic Frontier Foundation warned Congress that federal agencies must establish robust civil liberties protections before deploying artificial intelligence systems, particularly for surveillance purposes.

Dr. Matthew Guariglia, EFF's Senior Policy Analyst, delivered testimony to the House Homeland Security Subcommittee on Cybersecurity and Infrastructure Protection, focusing on how generative AI could amplify unconstitutional government overreach. His appearance came during a hearing examining frontier AI models, agentic systems, and their implications for cybersecurity and critical infrastructure.

Guariglia emphasized that generative AI applied to mass surveillance would "supercharge unconstitutional violations of civil liberties." The combination of government secrecy and proprietary commercial AI systems creates a dangerous accountability gap, preventing both the public and lawmakers from understanding when these systems fail.

Why it matters

As federal agencies rapidly adopt AI tools, the lack of transparency around their deployment creates blind spots in oversight. When AI systems make consequential errors in government contexts—from cybersecurity vulnerabilities to impacts on individual rights—classification and corporate secrecy can hide these failures from scrutiny. This testimony signals growing concern that the pace of government AI adoption is outstripping the development of necessary legal and procedural safeguards.

Track record of AI failures in government

The testimony highlighted documented instances where AI systems have produced serious errors in government applications. Guariglia cited examples ranging from false legal citations to a Department of Homeland Security incident where an AI mistake resulted in recruits being deployed without proper training.

"AI also has a track record of getting things wrong," Guariglia stated in his opening remarks. He noted that classification restrictions likely conceal additional consequential failures that would inform a more complete understanding of AI risks in government operations.

Focus on agency accountability

When questioned by Subcommittee Ranking Member Delia Ramirez (D-Ill.), Guariglia reframed the AI governance challenge. Rather than focusing solely on the technology itself, he argued the critical issue is controlling how government agencies deploy these systems.

"At this level the question is not how do we rein in AI, it's how do we rein in the agencies that would unleash AI on the American public," he said.

This perspective shifts the regulatory conversation from technical AI safety to institutional accountability and the preservation of constitutional protections in an era of powerful new surveillance capabilities.

The full prepared testimony is available through the Electronic Frontier Foundation, which first reported these details.