SweetREX AI: US Government's New Deregulation Tool
A new artificial intelligence tool, dubbed SweetREX Deregulation AI, is at the forefront of the US government’s ambitious drive to dismantle federal regulations. This initiative, spearheaded by the Office of the Chief Information Officer at the Office of Management and Budget (OMB), aims to streamline the regulatory landscape by identifying and eliminating rules not explicitly mandated by statute. The development and deployment of SweetREX, formally known as the SweetREX Deregulation AI Plan Builder (SweetREX DAIP), align directly with the Trump administration’s “Unleashing Prosperity Through Deregulation” executive order and the broader objectives of Project 2025, a blueprint for a potential second Trump term that prioritizes industrial-scale deregulation. Indeed, the nascent Department of Government Efficiency (DOGE), a key player in this effort, has projected that half of all federal regulations could be scrapped, a claim highlighted in a July 2025 internal presentation.
SweetREX was conceived by associates of DOGE, operating from within the Department of Housing and Urban Development (HUD), with plans for its eventual rollout across numerous US agencies. A recent video conference, attended by staff from agencies including the Environmental Protection Agency, the Department of State, and the Federal Deposit Insurance Corporation, offered a glimpse into the tool’s capabilities. Co-leading this crucial discussion was Christopher Sweet, a DOGE affiliate and the tool’s primary developer, whose name inspired the project. Sweet, who was recently a third-year student at the University of Chicago and initially introduced as a “special assistant,” indicated that federal workers would increasingly leverage tools from major AI developers like Anthropic and OpenAI, but confirmed that SweetREX itself primarily uses Google’s Gemini family of models.
The functionality of SweetREX centers on its “decision tool” feature, designed to flag sections of regulations it deems superfluous based on relevant statutes. Government attorneys and policymakers are then tasked with reviewing the AI’s proposed amendments and making necessary adjustments. The system also generates a draft of the AI-modified regulation for further scrutiny. Beyond identifying redundant clauses, SweetREX is designed to categorize public comments on proposed regulatory changes into AI-determined “buckets” and even differentiate between individual and “sophisticated” corporate commenters. Scott Langmack, a DOGE-affiliated senior adviser at HUD and COO of technology firm Kukun, co-led the call with Sweet. Langmack, responsible for demonstrating the tool and pitching its benefits to various agencies, asserted that SweetREX could dramatically cut the time required for reviewing and proposing regulatory edits, from months to mere hours or days.
Despite these optimistic projections, the tool’s real-world application has drawn mixed reactions. While some participants on the recent call lauded SweetREX as “awesome” and “great,” a HUD worker previously described the process of scrutinizing AI-flagged regulations to determine their continued necessity as “painful.” Earlier reports on a similar AI tool at HUD revealed its capacity to quantify word count reductions in regulations and assign a “noncompliant” percentage, though the methodology for this calculation remained unclear.
The push for deregulation also intersects with high-profile figures. During the call, Steve Davis, a prominent lieutenant to Elon Musk who recently left his post as a special government employee with DOGE to return to his role as president of Musk’s Boring Company, inquired about the possibility of open-sourcing SweetREX. Davis’s previous involvement with DOGE had reportedly caused discomfort among staff, who perceived him as continuing to direct the department even after his official departure from government service. Neither Christopher Sweet, OMB, Google, Scott Langmack, nor Steve Davis provided immediate comments regarding the tool or its implications.