Your phone knows where you slept last night. Your email provider knows who you argued with this morning. Your browser knows what you’re worried about.
Individually, these data points are mundane. Assembled by an AI system with access to government surveillance databases? They become the most detailed profile of your life ever constructed — and right now, Congress is fighting over whether to let that happen without a warrant.
The FISA Fight Reaches Breaking Point
Section 702 of the Foreign Intelligence Surveillance Act lets the government collect communications of foreigners abroad without a warrant. When an American contacts someone overseas, their communications get swept up too. The government can then search those intercepted American communications — no warrant required.
This authority nearly expired on April 21. Speaker Mike Johnson attempted a five-year clean extension at midnight, only to watch 20 Republican Freedom Caucus members vote it down. A follow-up vote at 2:07 AM for an 18-month extension also failed. Congress punted with a 10-day extension while scrambling for a deal.
Johnson’s latest attempt: a three-year version with additional safeguards — but crucially without the warrant requirement that privacy advocates consider non-negotiable.
“The latest House FISA bill is a rubber stamp for Trump and Kash Patel to spy on Americans without a warrant,” Senator Ron Wyden said. “Don’t fall for fake reforms.”
Why AI Turns Surveillance Into Something Entirely New
For decades, the practical limit on mass surveillance was human attention. The NSA could collect everything, but analysts could only look at so much. Sheer volume acted as accidental privacy protection.
AI obliterates that limitation.
“Imagine instead of doing a query with one person that you turned AI loose on these databases,” Rep. Thomas Massie said this week. “There’s virtually nothing the government can’t know about you.”
LLMs could do the work of entire intelligence analyst teams — cross-referencing location data, communication patterns, financial records, and social media activity — in seconds rather than weeks. The cost drops from millions in analyst salaries to pennies in API calls.
Karen Levy, professor of information science at Cornell, nailed it: “A lot of what we think of as privacy protection isn’t something written in the law. It just has to do with how hard or how expensive it is to learn stuff about people.”
AI makes learning stuff about people essentially free.
The Data Broker Loophole
Here’s the absurd paradox enabling all of this. If police suspect you of a crime and want your phone’s location data, they need a warrant. Fourth Amendment, full stop.
But if the government buys that same location data from a data broker — a company that hoovered it up from your apps — no warrant required. The government isn’t “searching” you; it’s purchasing commercially available information.
A 2019 New York Times investigation showed that “anonymized” location data could be de-anonymized by identifying where a phone spent its nights and days. That took journalists days. An LLM agent could do it for millions of records simultaneously.
The Trump administration’s AI policy framework actually urges Congress to “allow industry and academia to use federal datasets to train AI” — a provision privacy advocates view as an invitation to feed surveillance data into machine learning systems.
The Corporate Split
This story has a revealing corporate dimension. Anthropic refused to let its AI models analyze commercially available data on U.S. citizens for the Department of Defense. CEO Dario Amodei called AI-enabled mass surveillance something that “could constitute a crime against humanity.”
OpenAI agreed to a DOD deal hours later. Faced immediate backlash. Revised the terms.
The government doesn’t need to build its own AI. It just needs a willing vendor.
Strange Bedfellows Fighting Back
What’s unusual is the political alignment. The Freedom Caucus is breaking ranks with the Trump administration. Democrats like Jamie Raskin are joining Republican libertarians like Massie and Warren Davidson. Boebert and Massie introduced the Surveillance Accountability Act to close data-collection loopholes.
“Times have changed since 2024. The watchdogs are gone,” Raskin said, pointing to the administration’s hollowing out of oversight mechanisms like the Privacy and Civil Liberties Oversight Board.
The CIA counters that Section 702 “is the most extensively overseen US intelligence collection tool” and helped prevent a terror attack at a Taylor Swift concert in Austria.
What’s Actually at Stake
The practical implications of AI-powered surveillance:
- Location tracking at scale — correlating phone movements with known locations to build behavioral profiles for millions simultaneously
- Communication analysis — LLMs reading and summarizing thousands of intercepted emails without human review
- Pattern recognition — identifying social networks and predicting behavior across entire populations
- Data fusion — combining purchased broker data with intercepted communications for comprehensive profiles no single source could provide
DHS has already served hundreds of subpoenas to Google, Reddit, Discord, and Meta requesting data on users discussing immigration policy online. Add AI to that collection capability and you have something resembling the social credit system Americans love to criticize when China does it.
The Clock Is Ticking
The 10-day extension expires imminently. Even if Section 702 lapses, existing surveillance continues through March 2027 — this isn’t a clean kill switch.
But the larger question transcends one law. The entire legal framework governing surveillance was designed when data collection was expensive and analysis was slow. AI makes both trivially cheap. The laws we write in the next few weeks will determine whether AI becomes a tool for protecting citizens or profiling them.
The bipartisan opposition is encouraging. The institutional forces pushing for expanded surveillance — intelligence community, White House, willing tech vendors — are powerful. Place your bets accordingly.