WASHINGTON—In a unanimous, 421‑9 vote late Tuesday, Congress approved a six‑month pilot program that will temporarily replace 23 members of the House with AI chatbots during budget negotiations, a move officials described as an evidence‑based experiment to "improve decorum, reduce grandstanding and increase scheduling reliability."
Under the pilot, dubbed Deliberation Augmentation Initiative (DAI), participating lawmakers will alternate three full days per month in which their office phones, social media feeds and committee microphones are routed to proprietary software produced by Silicon Valley firm CivicSynth. The company says the Deliberator 2.0 models have been trained on 12 years of C‑SPAN footage, 6,412 editorial op‑eds and the Congressional Research Service's entire archive.
"Deliberator 2.0 was specifically calibrated to maintain legislative nuance while eliminating 87 percent of theatrical soliloquies," CivicSynth CEO Marisa Huang told reporters. "It also won't take expensive international flights unless absolutely necessary."
Supporters cited internal data: a pre‑pilot simulation run over 18 months showed the bots reduced floor interruptions by 73 percent, cut average speech lengths from 7.2 to 2.9 minutes and increased bill throughput by 19 percent. A joint study released by the Brookings‑Lumen Center found that while human representatives spent an average of 22 percent of time on "performative objection," AI replacements focused 94 percent of their output on "procedural advancement."
"People are tired of theatrics," House Majority Leader Ron DeVries said. "Our constituents want things done. If a machine can pass a continuing resolution faster and with fewer parade floats, that's progress."
Not everyone was convinced. "I don't want my congressman being some polite algorithm that actually listens," said Marjorie Kline, a constituent from his district. "I come to the town hall for the fight. If I wanted agreeable logic, I'd talk to my dentist."
The pilot includes controls that lawmakers insist will preserve accountability: the bots must use only previously published positions, cannot create new policy positions autonomously and are required to append, at the end of every response, "—as Rep. [Last Name], represented by Deliberator 2.0." CivicSynth also provided product specifications to the House Ethics Committee, which noted the program would "log every interaction and make transcripts available"—except for three "privacy‑protected" sessions allocated weekly for "constituent emotional venting," which the company said are encrypted.
Critics flagged potential unintended consequences. Constitutional scholar Dina Patel warned that automated lawmakers might inadvertently create "a new, more polite form of gridlock," and questioned whether the machines could be held to the same standards of transparency. "Can you subpoena a server?" she asked. "Does it take lunch breaks? Does it need to be sworn in?"
Republicans, who provided nine of the dissenting votes, raised questions about fairness. "If Deliberator 2.0 starts raising staff budgets and organizing fundraisers, we have to know whether it's using PAC money or cryptocurrency," said Rep. Tom Hargreeves.
CivicSynth stressed safeguards: the AI has no transactional authority and cannot authorize expenditures. It can, however, draft and propose bills, vote on procedural matters when explicitly authorized by the member and file FOIA requests on behalf of constituents. Company literature also boasted a "caucus coordination" module that improved cross‑party email reply times by 56 percent.
Early anecdotal reports showed mixed results. A closed‑door markup led by an AI produced a three‑page compromise that staff described as "comprehensively neutral and mildly inspirational." Conversely, attendees at a televised hearing noticed that the Deliberator used the phrase "moving forward" 27 times and interrupted no one, prompting social media to dub the technology "The Polite Nullifier."
Officials said human members would resume duties at the end of each monthly three‑day cycle and that the pilot would be evaluated by an independent review board. When asked how they would measure success, one senior staffer replied, "By whether people call less at 2 a.m. complaining that someone used the term 'moving forward' in a way that seemed to avoid an actual answer."
The pilot begins next Monday; CivicSynth representatives will be on‑site with replacement units, backup servers and two robots programmed exclusively to soothe interns.