S-4402-119
Read twice and referred to the Committee on the Judiciary.
Sponsored by Adam Schiff (D-CA)
What it does
This bill would require the Attorney General and the Director of National Intelligence to jointly submit a report — within 120 days of enactment — describing all current uses of artificial intelligence (AI) systems that have access to "unminimized" information collected under the Foreign Intelligence Surveillance Act (FISA). Unminimized information is raw surveillance data that has not yet had identifying details about U.S. persons removed. The report would detail each AI system's name, purpose, training data, developer, access scope, and any related Foreign Intelligence Surveillance Court (FISC) orders. Going forward, the bill would also require the Attorney General and DNI to notify Congress and the FISC before granting any new AI system access to unminimized FISA data, along with a legal compliance assessment.
Who benefits
U.S. persons whose communications may be incidentally collected under FISA and processed by AI systems without their knowledge. Congressional oversight committees (Senate and House Judiciary, and the intelligence committees) that would gain visibility into previously undisclosed AI use. Civil liberties organizations and researchers focused on surveillance transparency. The FISC judges who would receive disclosures and could issue orders governing AI use. Journalists and the public, who would have access to the unclassified version of the report.
Who is hurt
Intelligence agencies (FBI, NSA, CIA, and others) that may face operational constraints or disclosure burdens on existing AI programs. Private technology contractors who developed or trained AI systems used in intelligence work, whose methods and data practices would be disclosed to Congress and the FISC. Counterintelligence and national security operations that rely on speed and secrecy, which could be slowed by pre-deployment notification requirements. Agencies may also face resource costs to compile and maintain the required documentation.
Supporters argue
Supporters argue that AI systems processing raw, unminimized FISA data — which can include the communications of U.S. persons — represent a qualitatively new surveillance capability that existing oversight frameworks were never designed to address. They contend that Congress and the FISC currently have no systematic visibility into which AI tools access this sensitive data, how they were trained, or whether they comply with FISA's minimization requirements. Without this baseline inventory, meaningful oversight is impossible, and the risk of unlawful or erroneous processing of Americans' private communications goes undetected and uncorrected.
Opponents argue
Opponents argue that requiring detailed disclosures about AI systems — including their names, training data, developers, and access parameters — to multiple congressional committees and FISC judges creates significant risks of intelligence source and method exposure, even in classified form. They contend that the pre-deployment notification requirement for future AI systems could slow time-sensitive national security operations and that existing internal oversight mechanisms within the intelligence community, including the Privacy and Civil Liberties Oversight Board and Inspector General offices, already provide adequate checks without the operational risks of broader disclosure.