Overview
- A screenshot circulated showing an amendment summary for the 2027 National Defense Authorization Act with the line “11:25 AM????Claude responded,” revealing Anthropic’s Claude output in the public post.
- Rep. Anna Paulina Luna first acknowledged staff used AI to “correct a draft” then edited her statements to say Claude was used only to spell- and grammar-check the amendment summary and not the official bill text.
- All formal bill text in the House is produced by the Office of Legislative Counsel which lawmakers say is prohibited from using AI, and Luna has maintained that office drafted the amendment text.
- The posted summary was revised to remove the visible Claude reference after the screenshot spread and, as of the latest reporting, no formal investigation or procedural action has been announced.
- Commentators tied the episode on June 24 to a broader pattern of problematic AI use by public officials and legal actors and warned it could deepen demands for clearer rules on when congressional staff may use large language models.