AI could finally make congressional hearings work better

Sen. Bernie Sanders (I-Vt.) recently did something no member of Congress has done before. He sat down with Claude, the AI assistant developed by Anthropic, and conducted what amounted to a wide-ranging policy interview. He pressed Claude on health care, income inequality, corporate greed, and the future of democracy. 

The exchange, posted on YouTube, was remarkable not for what the AI said but for what it revealed: Generative artificial intelligence can engage substantively with complex policy questions more coherently than most witnesses who appear before congressional committees. That should be a wake-up call on Capitol Hill.

Anyone who has watched a congressional hearing in the past two decades, and especially in recent months, knows the ritual. Members of Congress use their allotted five minutes not to ask probing questions but to deliver mini speeches aimed at cable news clips and social media virality. The supposed witnesses, meanwhile, have been coached by armies of lawyers to say as little as possible, deflect with bureaucratic jargon, or simply run out the clock with filibuster-length non-answers. The result is political theater that generates heat but rarely light. The public learns nothing. Accountability is nowhere to be found.

AI will not fix the grandstanding. That is a human choice, and members of Congress have shown no appetite for surrendering their camera time. But AI could fundamentally transform the other side of the equation by improving the quality of the questions being asked and the difficulty of evading them.

Consider what would happen if senators and representatives used a tool like Claude to prepare for hearings. An AI assistant could analyze thousands of pages of depositions, regulatory filings, financial disclosures, and prior testimony in hours.

It could identify contradictions between what a witness said last year and what they are expected to say tomorrow. It could draft sequences of questions designed not as isolated gotcha moments but as logical chains where each answer narrows the witness’s room to maneuver. A well-constructed AI-assisted line of questioning would function less like a scattershot press conference and more like a skilled cross-examination.

Sanders’s conversation with Claude suggests another possibility, allowing him to demonstrate its application in practice. He should use it to help draft questions for all the Senate committees he serves on, and to anticipate likely answers to those questions. If you know how a witness is likely to dodge, you can design the follow-up before the hearing even begins. AI does not get flustered. It does not forget the thread. It can game out five moves ahead while Sanders is still reacting to the first evasion.

This is not hypothetical futurism. The technology exists today. What is missing is the institutional imagination to deploy it. Congressional staff members are overworked, underpaid, and expected to become overnight experts on everything from semiconductor supply chains to social media algorithms to the pricing structures of insulin. They prepare briefing books that members often barely skim before walking into a hearing room. AI would not replace their judgment, but it would radically augment their capacity to prepare members like Sanders, who actually want to hold witnesses accountable rather than simply perform for the cameras.

There are legitimate concerns, of course. AI-generated questions could be manipulated to serve partisan agendas rather than genuine oversight. Members could become lazy, outsourcing their thinking entirely. And there is the uncomfortable question of what it means when an algorithm is better at democratic accountability than the elected officials themselves.

But these risks pale against the status quo, which is a system where hearings have become so performative that most Americans tune them out entirely. When the chair of a major committee cannot get a straight answer from a tech CEO about data privacy or a pharmaceutical executive about drug pricing, the institution itself loses credibility. 

AI will not fix everything. But it could make evasion significantly harder and preparation dramatically deeper. That alone would be worth the experiment.

Sanders stumbled onto something important in his conversation with Claude. He demonstrated that AI can be a serious interlocutor on policy substance. The next step is obvious: bring that capability into congressional hearing rooms. Not as a witness, and certainly not as a replacement for democratic deliberation, but as the best-prepared legislator in the building, who never forgets a prior contradiction or runs out of meaningful follow-up questions.

Stuart N. Brotman is Digital Media Laureate and Distinguished Senior Fellow at The Media Institute. A former Visiting Professor of Entertainment and Media Law at Harvard Law School, he is the author of “Free Expression Under Fire: Defending Free Speech and Free Press Across the Political Spectrum.”

Copyright 2026 Nexstar Media Inc. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.

More Congress Blog - Technology News

Here are Iran’s 5 conditions for ending war after rejecting US ceasefire plan

Judge asked to take Trump name off Kennedy Center

Platner holds commanding lead over Mills in Maine Senate race: Poll

Hegseth: Military chaplains will no longer display rank

Trump disapproval highest in both terms: Fox News poll

Democrats worry centrists will undercut them on DHS shutdown amid TSA meltdown

Trump says Iran’s ‘present’ was oil ships through the Strait of Hormuz

Conservatives shoot down Senate off-ramp on SAVE America Act

Group floats ‘Medicare by Choice’ as a Democratic health care alternative

Cruz asks Senate to hold his salary during DHS shutdown

Senate GOP-backed deal to end DHS shutdown draws extensive fire

Rogan: Democrats were tougher on border than Trump

Murphy on ‘$1.5 BILLION’ stock trade before Trump Iran ...

ICE officers now checking IDs in airport security lines

Johnson says supplemental Pentagon funds ‘probably’ in reconciliation as ...

Live updates: Trump says Iran ‘begging to make a deal’; DHS shutdown grinds ...

Navarro: Supreme Court ruling on tariffs ‘best possible outcome’ for ...

Democrats face test as rare Ethics hearing examines Cherfilus-McCormick fraud ...

© The Hill