The US is discussing using Google Gemini to draft federal regulations, starting with the Department of Transportation. The idea has already drawn criticism due to the risk of errors in documents that directly affect human safety.
The US is discussing using Google Gemini to draft federal regulations, starting with the Department of Transportation. The idea has already drawn criticism due to the risk of errors in documents that directly affect human safety.
According to Engadget, the pilot initiative is planned to be launched at the Department of Transportation (DOT). It was presented to employees within the department last month. In a letter to colleagues, DOT lawyer Daniel Cohen described AI as a tool that could significantly change the rulemaking process. DOT General Counsel Gregory Zerzan stated during the discussions that Donald Trump supports this direction and that DOT wants to become the first federal agency to be fully equipped with the ability to use AI to prepare regulatory texts. Later, according to the plan, the practice can be expanded to other departments.
The key argument for supporters is speed. Transcripts of the meeting obtained by journalists show that DOT management is eager to get a draft of the rule very quickly. Zerzan, according to media reports, said that the agency does not need a perfect text, but rather a «good enough» version, and that the agency wants to drastically reduce the time it takes to prepare documents. One of the employees who presented the program allegedly stated that a significant part of such documents consists of formal phrases, so AI can handle it. The material also mentions that the DOT has already used AI to prepare an unpublished Federal Aviation Administration rule.
Critics warn that the cost of getting it wrong here is too high. The DOT sets requirements for commercial aviation, hazardous materials regulations and driver qualifications, which means documents affect transportation safety every day. Former DOT acting chief AI officer Mike Gorton compared the approach to having an intern write key texts. Bridget Dooling, a professor at Ohio State University who studies administrative law, stressed that just because a model can generate a lot of words doesn’t mean it will be a good decision for the state. Additional context here is that the DOT has cut more than 4,000 employees since the start of Trump’s second term, including more than 100 lawyers, which could have pushed the agency to find a way to work faster.
US federal agencies have already used AI, but mostly as an auxiliary tool: for document translation, data analysis, and sorting through public comments. Using a generative model to write regulatory texts raises issues of accountability and quality: in such documents, wording is important, because it determines whether a rule will withstand legal challenge and how it will be implemented in practice.
Previously, dev.ua wrote about how Donald Trump announced that he would sign an executive order regarding the AI approval process to avoid different rules in each US state.
