Rep. Anna Paulina Luna, R-Fla., on Wednesday defended her staff’s use of the AI chatbot Claude to help summarize an amendment to congressional legislation, claiming it was a widespread practice after screenshots circulated online showing a summary of the amendment mistakenly included a reference to the LLM.
Rep. Anna Paulina Luna, R-Fla., leaves the U.S. Capitol on January 15.
CQ-Roll Call, Inc via Getty Images
Luna submitted an amendment to the National Defense Authorization Act, which approves various policies and expenditures for the Department of Defense, and screenshots published online suggested the summary of the amendment had been copied and pasted from Claude.
“11:25 AM????Claude responded:” the original amendment summary read before launching into what the amendment said, as first posted online by Washington Post journalist John Hudson.
The website has since been updated to remove the reference to Claude, though it is still visible through a Google search.
The congresswoman defended the AI usage in a post on X, noting no AI was used to draft the actual legislation, and the screenshot referencing Claude is only an “AI summary of the bill,” adding, “c’mon man.”
All legislative text in Congress is drafted by the House’s Office of Legislative Counsel, which Luna noted is prohibited from using AI, so no AI tools were used in the drafting of the bill itself—only the summary of what the amendment said.
Luna also confirmed her staff’s use of Claude in a post on X that’s since been deleted, writing her staff used Claude to edit the summary and mistakenly “didn’t edit,” adding “most” congressional staff use the AI tool and she told her staff “to make sure they are double checking and more thorough.”
Luna’s amendment to the bill, if it’s approved, would direct Defense Department resources to the southwest border of the U.S. No other lawmakers have yet commented on the apparent mistake in the amendment summary or whether their own staffers use AI to draft legislative summaries, as Luna claimed Wednesday.
As AI usage becomes more widespread, its use in government has become a growing issue, though the apparent mistake in Luna’s amendment summary is so far a rarity on Capitol Hill. A number of governments around the world have admitted to using AI to improve their workflows, such as in the U.K., where officials touted the use of an in-house AI tool to review responses to a government review, and in Italy, whose Senate uses AI to help manage amendments. In the U.S., ProPublica reported in January the Department of Transportation was planning to use AI to draft its regulations, quoting general counsel Gregory Zerzan as saying, “We don’t need the perfect rule on XYZ. We don’t even need a very good rule on XYZ … We want good enough.” Beyond government rulemaking, the use of AI has repeatedly been problematic in the legal sphere. Numerous courts have sanctioned attorneys for filing documents with false information hallucinated by AI, such as citing court cases that don’t exist. Michael Cohen, the former personal attorney to President Donald Trump, came under scrutiny in 2023 after his lawyers submitted filings with false court cases, with Cohen claiming he did research using Google’s Bard tool that he sent to his attorneys, and was unaware it was giving him false information.
Government by AI? Trump Administration Plans to Write Regulations Using Artificial Intelligence (ProPublica)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。