Congresswoman Denies AI Wrote Defense Funding Amendment After Claude Response Found Embedded in Bill Text
A political firestorm erupted on Capitol Hill after screenshots circulating on social media appeared to show that an artificial intelligence chatbot — specifically Anthropic's Claude — had been used to draft or summarize a congressional amendment tied to the 2027 National Defense Authorization Act (NDAA). Representative Anna Paulina Luna (R-FL) quickly pushed back against the accusations, insisting that while her staff may have used AI as a spellcheck tool, no legislation in her office is ever written by an AI system.
The incident has reignited a fierce debate about the role of artificial intelligence in American lawmaking — raising serious questions about transparency, accountability, and the future of democratic governance in the age of generative AI.
What Exactly Was Found in the Amendment Summary?
The controversy began when users on X (formerly Twitter) began sharing screenshots of an amendment summary related to the 2027 National Defense Authorization Act. Embedded within the document — a high-stakes piece of defense legislation that governs hundreds of billions of dollars in military spending — was a line that appeared to be a raw output from an AI chatbot conversation.
The text in question read: "Identical to H.R. 100 (118th Congress).11:25 AM????Claude responded: Requires the Secretary of Defense to designate Department of Defense activities, support, and operations at the southwest land border as a named operation with…"
The inclusion of a phrase like "Claude responded:" is a telltale sign that someone copied and pasted text directly from a conversation with Anthropic's Claude AI assistant — without properly cleaning up the output before submitting it as part of an official legislative document. Whether this was the result of carelessness, expediency, or a broader systemic practice remains a matter of public debate.
Rep. Luna's Official Response
Representative Luna moved quickly to get ahead of the story. In a statement, she acknowledged that her staff had used AI tools, but only in a limited and — in her framing — entirely benign capacity. According to Luna, the AI was used solely for "spellcheck" purposes in the amendment summary. She was emphatic that the bill text itself was not generated by any AI tool and reiterated that her office follows a firm policy: "NO Legislation is ever drafted with AI."
Luna's initial response, however, appeared somewhat contradictory to observers, as the visible AI output seemed to go well beyond a simple grammar or spelling correction. The presence of the phrase "Claude responded:" in the document suggests that at minimum, substantive content from an AI conversation was copied into an official government document — a distinction that many critics argue matters greatly, regardless of whether it technically constitutes "drafting."
Why This Incident Matters for American Democracy
The stakes here go far beyond one congresswoman's office. The NDAA is one of the most consequential pieces of legislation passed by Congress each year. It authorizes funding for the entire United States military, sets defense policy priorities, and addresses national security issues ranging from cybersecurity to border operations. An error — or worse, an undisclosed AI-generated passage — embedded in such a document carries enormous implications.
- Accountability: Voters and oversight bodies have a right to know how legislation is being drafted. If AI tools are being used without disclosure, it undermines the transparency that democratic processes depend on.
- Accuracy: AI language models, including Claude and others, can generate plausible-sounding but factually incorrect information. In a defense context, such errors could have serious real-world consequences.
- Legal integrity: Legislative language must be precise. AI-generated text, while often fluent, may introduce ambiguities that create loopholes or unintended legal interpretations.
- Public trust: When citizens discover that AI may be writing the laws that govern them — even partially — it erodes confidence in their elected representatives.
AI in Congress: A Growing and Largely Unregulated Reality
The Luna incident is not occurring in a vacuum. The use of AI tools on Capitol Hill has been quietly expanding for years. Staffers use AI to summarize lengthy reports, draft constituent correspondence, and research policy positions. Some offices have begun experimenting with AI for more substantive tasks, including analyzing bill language and drafting initial policy memos.
Yet Congress has no formal, binding rules about how or when AI can be used in the legislative process. The Congressional Research Service has issued guidance, and individual offices have developed their own informal policies, but there is no chamber-wide standard governing AI disclosure or usage limits. This regulatory gap means that what happened in Luna's office — whether the result of a genuine mistake or something more deliberate — could be happening in dozens of other offices without anyone knowing.
Some lawmakers have called for clearer rules. Proposals have circulated that would require any AI-assisted legislative language to be disclosed, similar to the way lobbyist contributions or financial conflicts of interest must be reported. So far, none of these proposals have advanced into formal policy.
The Broader AI Transparency Debate
This episode taps into a wider societal conversation about AI transparency that extends well beyond government. In journalism, academia, law, and medicine, institutions are grappling with the same core question: when AI is used to produce or assist with professional work product, should that use be disclosed? And if so, how?
For elected officials, the argument for disclosure is arguably stronger than in almost any other field. Legislators are hired by the public to represent public interests. Their work product — the laws they write — directly governs the lives of millions of people. If constituents cannot know whether their representative's legislative staff wrote a bill or whether an AI chatbot did, something fundamental about representative democracy is compromised.
What Should Congress Do Next?
Experts in tech policy and government accountability have begun calling for a series of concrete steps in the wake of the Luna controversy. At a minimum, many argue that Congress should establish clear disclosure requirements when AI tools are used in any part of the legislative drafting process. Beyond that, some advocate for the creation of official AI usage guidelines developed in partnership with technology experts and ethics scholars.
There is also a training dimension. The Luna incident suggests that even when staff use AI tools only in a supplementary capacity, they may not be adequately trained to handle AI outputs responsibly. Copying raw chatbot responses — complete with timestamps and system labels like "Claude responded:" — into official documents is exactly the kind of error that proper training could prevent.
Conclusion: A Wake-Up Call for AI Governance on Capitol Hill
Whether Rep. Luna's staff used AI to write the defense amendment or merely made a sloppy copy-paste error, the result is the same: a fragment of an AI chatbot conversation ended up in an official congressional document tied to one of the most important defense bills of the year. That alone should be a wake-up call.
The question is no longer whether AI will play a role in how laws are made in America — it already does. The question is whether Congress will have the self-awareness and political will to govern that role before it shapes legislation in ways the public never consented to and may never even know about.
For now, all eyes remain on Capitol Hill — and on the AI tools quietly running in the background of American democracy.