Congresswoman Luna denies that Claude wrote a US defense budget amendment

Congresswoman Luna Denies That Claude Wrote Amendment to U.S. Defense Budget

Official documents for a U.S. defense budget amendment contained an artifact of a conversation with Claude — Congresswoman Anna-Paulina Luna insists that her staff used AI only for text checking, not for drafting legislation.

What ended up in the screenshot

In a summary of the amendment to the 2027 National Defense Authorization Act — NDAA, the main annual U.S. military budget — several X users discovered something unexpected. Among the official text was the phrase: "11:25 AM ???? Claude responded:" — followed by a brief summary of the amendment's requirements regarding giving operations of the Department of Defense on the southwestern border the status of an officially named operation. The screenshot spread across X instantly. The conclusion seemed obvious: someone from the congresswoman's staff copied an answer from a Claude chat directly into an official document and failed to delete the dialog header. Criticism followed immediately.

What Luna responded

The Republican congresswoman from Florida reacted quickly. According to her, Claude was used exclusively for "spell-checking" in the brief description of the amendment — but not in the legislative text itself.

"No bill is ever drafted with AI," Luna wrote.

The initial response, however, read ambiguously — it seemed as though AI wasn't used at all. Luna had to clarify her position separately: she was referring only to the supplementary summary, not the amendment text itself. The difference is fundamental. An amendment summary is an explanatory document read by journalists, lobbyists, and fellow lawmakers. It was precisely in this document that the Claude trace appeared. According to the congresswoman, the actual text of the amendment was written by people only.

The big problem of a small slip-up

The incident exposes a systemic problem: the U.S. Congress still has no unified policy for using AI in legislative work. The situation in Luna's office is a symptom, not an exception.

• No common standard exists for using AI tools in drafting bills
• Some offices have implemented their own bans on external chatbots when preparing drafts
• Amendment summaries end up in public documents and shape the perception of legislative norms
• A dialog artifact in an official document points to weak control of work processes
• The NDAA annually collects a huge number of amendments, creating significant pressure on staff

The pressure of tight deadlines pushes staff to use any available tools — including AI — without clear protocols and without understanding that "raw" output might end up in a public document.

What this means

The incident in Luna's office is less of a political scandal than a reminder of operational risks from unstructured AI use. When there are no internal rules, the phrase "Claude responded:" easily ends up in official paperwork. Congress will have to develop a unified position — a ban, mandatory disclosure, or standards for acceptable use. Until that happens, such cases will repeat.