The screenshot lands in your feed: a line of legislative text with the phrase “Claude responded” sitting awkwardly in the middle. You scroll back and forth, trying to figure out whether this is a prank, a leak, or a mistake. The room gets quieter—I felt that shift the moment the image blew past my timeline.
I’m going to walk you through what happened and why it matters. You already know the players—Rep. Anna Paulina Luna, Anthropic’s Claude, Grok—and you’ll see how a single line can change a narrative overnight.
A screenshot of draft text went viral on X.
The image showed a bill summary for the 2027 National Defense Authorization Act with the phrase “Claude responded” embedded inside. That line turned the file into a headline before anyone verified the file’s provenance.
Rep. Anna Paulina Luna (R-FL) Used Anthropic’s Claude AI Chatbot to Draft Amendment to Defense Bill pic.twitter.com/Ls9oK26M8W
— NewsWire (@NewsWire_US) June 24, 2026
The screenshot suggested that everything was copied wholesale from an Anthropic Claude chat session. The screenshot was a smoking gun, and once a smoking gun appears online, the story acquires gravity.
Luna initially appeared to admit AI had been used and then reworded her response.
Her first tweet read like a candid moment: “Yeah my staff used AI to correct a draft text and didn’t edit.” That tweet vanished, replaced by a narrower claim that Claude was used only to spell- and grammar-check an amendment summary, not the amendment language itself.
I watched the edits happen in real time. You can see the shift: what started as an apparent acceptance moved into a more defensive, precise denial.
Did Rep. Luna use AI to write the bill?
Short answer: the screenshot strongly suggests staff used Claude to draft or summarize text, and Luna’s deleted tweet reads like an admission. Her follow-up emphasized that official bill text comes from the House Legislative Counsel, which is barred from using AI—but that doesn’t erase the screenshot.
Can AI draft legislation?
Legally, official bill text in the House typically comes from nonpartisan lawyers at the Legislative Counsel; they’re not supposed to run drafts through chatbots. But staff can write summaries, memos, and amendment descriptions—and those are precisely the places AI is already being used. If you work around the edges, AI can influence what lawmakers see before they submit anything to counsel.
People on both sides reacted with instinctive distrust.
Pew Research finds only about 16% of American adults say AI will have a positive effect on society, so skepticism is the norm. When you tell someone that chatbots touched a defense bill—even a summary—the response is visceral.
That’s where AI became a scapegoat: it’s easier to blame the bot than to explain staffing habits or rushed workflows. I’ll be blunt—you should be skeptical when important language is produced without clear human review.
Luna’s persona sharpened the story.
She’s a familiar figure: a staunch Trump ally who has traded the Hill for loud media moments, appearing on Joe Rogan’s podcast and advancing combative messaging on X. Her offhand line—“Btw love Claude but Grok is way more savage”—was part deflection, part culture-war signaling.
If you follow Congressional theater, this is classic playbook: admit a small part, reframe the issue as a media stunt, and then cast doubt on the motives of the person who posted the screenshot.
The stakes are not only procedural; they’re perceptual.
When a member of Congress is linked to AI drafting—even indirectly—public trust frays. You and I both know that lawmaking is a trust game. If important policy language is seen as machine-made, people start asking whether decisions are being outsourced.
Is it common for staff to use AI for legislative work?
Yes—staff across parties use chatbots for summarizing, proofreading, and brainstorming. Tools like Anthropic’s Claude and the Grok model on X are part of the daily toolkit. That doesn’t mean the official house text was machine-written, but it does mean machine output is seeping into the process.
Here’s what I’d watch next: requests for audit trails from Legislative Counsel, transparency about staff workflows, and whether offices update rules about AI use. If you’re interested in policy integrity, push for those traces—otherwise we end up arguing about a line in a screenshot instead of fixing the underlying process.
This episode feels small but it matters—the screenshot sparked a wider debate about power, trust, and the role of AI in lawmaking. Do you want machines shaping the language that governs your life?