AI in Governance: Rep. Luna Denies Using Claude for Defense Bill Text

The intersection of generative AI and legislative drafting has come under intense scrutiny following a controversy involving Florida Representative Anna Paulina Luna. After screenshots surfaced suggesting Anthropic’s Claude was involved in drafting defense funding details, Luna clarified the role of AI in her office's workflow.

The Claude Controversy and the NDAA Amendment

The controversy erupted on X (formerly Twitter) when users shared screenshots of an amendment summary related to the 2027 National Defense Authorization Act (NDAA). The leaked text contained a peculiar formatting artifact: "Claude responded: Requires the Secretary of Defense to designate Department of Defense activities..." This specific phrasing strongly suggested that a prompt had been processed through Anthropic's Claude LLM to generate the descriptive summary.

Initial reactions from Rep. Luna suggested that staff had used AI to correct draft text without further editing. However, as public speculation grew regarding whether AI was being used to write actual law, the Congresswoman issued a corrective statement. She clarified that the AI was utilized solely for "spell/grammar check" on the amendment's summary, rather than the legal text of the amendment itself.

Distinguishing Between Summary and Legislation

A critical distinction in this debate is the difference between an "amendment summary"—an explanatory tool used for quick reading—and the "actual amendment text," which carries legal weight. Luna emphasized that no actual legislation is drafted with AI in her office, noting a vital institutional safeguard.

According to Luna, all official bill text from the House originates from the House Legislative Council, an entity that is strictly prohibited from using AI tools. This distinction is vital for maintaining the integrity of legal language, where a single hallucination or linguistic nuance error from an LLM could lead to significant unintended consequences in national defense policy.

The Growing Trend of AI in Public Policy

This incident is not an isolated case of "AI leakage" in professional environments. The broader AI landscape is seeing a surge in unauthorized or unvetted AI usage across various governance sectors:

  • Legal Misconduct: Judges have repeatedly caught lawyers using chatbots to draft legal filings that include fabricated citations.
  • Municipal Errors: City officials in Brazil recently approved an ordinance that was unknowingly written by ChatGPT.
  • State-Level Usage: Arizona state representative Alexander Kolodin has openly admitted to using ChatGPT to assist in writing state-level legislation.

As LLMs like Claude and ChatGPT become ubiquitous in the workplace, the risk of "shadow AI"—where employees use unapproved tools for official tasks—poses a significant challenge to the accuracy and accountability of public institutions.

Key Takeaways

  • Clarified Usage: Rep. Anna Paulina Luna denied using AI to draft defense legislation, asserting that Anthropic's Claude was only used to spellcheck and summarize an amendment.
  • Institutional Safeguards: Official House bill text is drafted by the House Legislative Council, which is prohibited from using AI to ensure legal precision.
  • Systemic Risk: The incident highlights a growing trend of unvetted AI integration in law and governance, ranging from legal filings to municipal ordinances.