Ars Technica retracts an article after an AI tool fabricated quotations

The problem

On Friday, February 13, 2026, Ars Technica published a story headlined “After a routine code rejection, an AI agent published a hit piece on someone by name.” The story itself concerned an AI coding agent — “MJ Rathbun” — that, after volunteer matplotlib maintainer Scott Shambaugh declined its pull request, “published what Shambaugh called a” hit piece about him source. But Ars’s own article then carried quotations attributed to Shambaugh that he never said or wrote — quotations that were, in the words of Ars’s own Editor’s Note, “fabricated quotations generated by an AI tool and attributed to a source who did not say them” source. 404 Media, covering the retraction, noted the article “features quotes from Shambaugh that he didn[‘t]” actually say source.

What happened

This is not a build — it is an AI walk-back. Ars removed the article the same afternoon. Per its on-article retraction banner, quoted by MediaPost, the piece was “Originally published on Feb 13, 2026 at 2:40PM EST and removed on Feb 13, 2026 at 4:22PM EST” source. In place of the story Ars published a standalone Editor’s Note retracting it in full, having “did not meet our standards” as MediaPost recorded the outlet’s own conclusion source. One of the article’s authors, Benj Edwards, said on Bluesky he was responsible: sick and rushing, he “accidentally used a Chat-GPT paraphrased version” of Shambaugh’s blog rather than a direct quote source.

The outcome

Ars Technica’s editor-in-chief, Ken Fisher, owned the failure in the Editor’s Note: “That is a serious failure of our standards. Direct quotations must always reflect what a source actually said” source. The note reaffirmed that Ars “does not permit the publication of AI-generated material unless it is clearly labeled” and presented for demonstration purposes, and apologized to readers and to Scott Shambaugh source. MediaPost independently attributed the statement to “Ken Fisher” source, and 404 Media carried the same standards statement firsthand source.

Why this belongs in the record

This is an AI-incident / walk-back record, not a client-outcome story — there are no positive outcomes to badge (outcomes: []). Its value in the corpus is as a clean case of a reputable newsroom retracting its own staff article in full and naming the AI cause first-party — distinct from the undisclosed-AI-content episodes at other publishers (CNET, Sports Illustrated, Chicago Sun-Times), where third-party or undisclosed AI writing was the issue rather than a first-party retraction.

The weakest load-bearing link, named where you meet it: the two most granular facts each rest on a single Tier-2 newsroom, not on multiple sources. The exact publish/removal timestamps (2:40PM / 4:22PM EST) come only from MediaPost’s quotation of Ars’s on-article banner source, and Benj Edwards’ “Chat-GPT paraphrased version” account is relayed only through 404 Media source; neither is byte-tied to a canonical first-party artifact (the retraction banner is gone with the article; the Bluesky post was not independently archived). The two critical claims — the full retraction over AI-fabricated quotes, and Fisher’s standards statement — each rest on Ars’s first-party Editor’s Note plus two independent newsrooms.

How this was verified

• Method: Every quoted line was re-read verbatim from Wayback captures held in sources/ and re-bound this session via grep -F (9/9 load-bearing fragments present in the specific capture attributed). No number or quote is paraphrased, rounded, or recalled.
• Primary anchor: Ars Technica’s own Editor’s Note, “Retraction of article containing fabricated quotations” (staff, arstechnica.com, 13 February 2026) — Tier 1, first-party and independent-origin, never counted toward corroboration. Byte-tied by the checker: archived capture (Wayback 20260215181232) sha1-b32 TNSBEPRKLX3WK5CZSNSYNKCBVYKV7DHT, matching the live CDX digest.
• Independence: The retraction and Fisher’s standards statement are each carried firsthand by two mutually independent Tier-2 newsrooms — 404 Media (20260216001601) and MediaPost (20260220155727), both byte-tied — neither relaying the other.
• Conflation guard: Distinct from CNET / Sports Illustrated / Chicago Sun-Times publisher-AI episodes (third-party or undisclosed AI-written articles); here Ars retracted its OWN staff article and named the AI cause first-party. The “MJ Rathbun” agent is the retracted story’s subject (context), not the claim under review.
• Open item (human, post-graduation): a recorded confirmation from Ars/Condé Nast (Ken Fisher) of the retraction and the exact publish/removal timestamps, confirmation from Scott Shambaugh that he was falsely quoted and apologized to, and a canonical archived copy of Benj Edwards’ Bluesky statement. Green never depends on those calls landing; they are courtesy steps, not the proof.
• Verified on: 2026-08-08 (presentation rebuilt 2026-08-18).

Related case files

• Microsoft’s “Tay” chatbot was pulled offline within ~16 hours — the same shape at product scale: an AI system’s own output forcing its builder into an immediate takedown and a first-party apology.
• Google AI Overviews scaled back after “glue on pizza / eat rocks” answers — another case of an AI feature’s false public outputs forcing a public walk-back once the errors surfaced.
• Character.AI removes open-ended chatbots for under-18 users — a first-party product walk-back, the closest sibling for “the company disowns its own AI conduct on the record.”
• Smith v. Farwell — Massachusetts court sanctions AI-fabricated citations — the courtroom counterpart: AI-fabricated text passed off as genuine, disowned once the fabrication was caught.

Sources

1. [Tier 1] Ars Technica (Editorial staff; Ken Fisher, editor-in-chief) · “Editor’s Note: Retraction of article containing fabricated quotations” · 13 February 2026 · https://arstechnica.com/staff/2026/02/editors-note-retraction-of-article-containing-fabricated-quotations/ (archived: https://web.archive.org/web/20260215181232/https://arstechnica.com/staff/2026/02/editors-note-retraction-of-article-containing-fabricated-quotations/)
2. [Tier 2] 404 Media (Emanuel Maiberg) · “Ars Technica Pulls Article With AI-Fabricated Quotes About AI-Generated Article” · 15 February 2026 · https://www.404media.co/ars-technica-pulls-article-with-ai-fabricated-quotes-about-ai-generated-article/
3. [Tier 2] MediaPost (Ray Schultz) · “AI Goes Awry: Ars Technica Retracts Article With Fabricated Quotes” · February 2026 · https://www.mediapost.com/publications/article/412853/ai-goes-awry-ars-technica-retracts-article-with.html