Documents recently released by the Lever indicate that officials in the Trump administration considered deploying an artificial intelligence program, developed by a protégé of Elon Musk, to facilitate a sweeping deregulation initiative. The AI tool, known as “SweetREX,” was designed to identify and eliminate regulations deemed burdensome to businesses, thereby supporting President Donald Trump’s pro-business agenda.
Details from the documents reveal that SweetREX was specifically programmed to target rules that imposed costs on private enterprises, limited innovation, or involved race-based classifications. The tool was reportedly capable of analyzing over a hundred thousand public comments in under thirty minutes, a feature touted as a significant efficiency improvement over traditional regulatory review processes.
“The documents reveal, for the first time, the shortcuts this AI tool takes when deciding whether a regulation is legally required and whether its burdens outweigh benefits to the public,” said Daniel McGrath, senior oversight counsel at the legal advocacy group Democracy Forward, which obtained the records through a Freedom of Information Act request. The extent to which SweetREX was actually utilized during its proposed government operations remains unclear, as does whether its outcomes were subjected to any form of evaluation.
The Department of Government Efficiency (DOGE), which Musk spearheaded from January to May 2025, claimed to have achieved $215 billion in government savings through actions including administrative downsizing and job eliminations. However, historical precedents raise concerns about the efficacy of AI in regulatory decision-making, which has often resulted in poor outcomes affecting lives.
SweetREX was presented to Department of Housing and Urban Development employees as an “AI solution for regulation extermination.” According to the pitch, the aim was to eliminate any regulatory provisions seen as overreach or imposing unnecessary burdens beyond existing congressional legislation. The documents outlined how SweetREX could not only pinpoint which statutes should be eliminated but also draft notices for proposed rule changes and organize myriad public comments related to those regulatory cuts.
Promoters claimed that employing SweetREX could reduce the time required for regulatory analysis from an average of thirty-six hours to less than three. The AI’s criteria for identifying regulations to discard included whether the rule raised constitutional concerns, involved unlawful delegation of legislative authority, imposed high costs on private interests, or treated individuals differently based on race.
Supporters of SweetREX argued that it would assist various agencies in adhering to presidential executive orders aimed at deregulation, including those targeting environmental regulations imposed on fossil fuel companies. The AI documents also suggested that the program could provide “evidence-backed flags [to] protect you in court,” further underscoring its intended role in the administration’s regulatory strategy.
The first indications of SweetREX’s development surfaced in August 2025, leading to a lawsuit filed by Democracy Forward in October 2025. The organization sought to compel federal agencies to disclose documents detailing how AI was utilized to further Trump’s deregulation objectives. The recently released records are a result of that legal challenge.
As discussions about the implications of AI in governance continue, the case of SweetREX highlights the potential risks and rewards of integrating such technology into regulatory frameworks. The balance between efficiency and accountability remains a key concern as policymakers grapple with the evolving landscape of artificial intelligence.
