10 responses to “Eye on AI 👁️2/12”

1. 

   granville

   02/13/2026

   Currently fascinated by this story right now, you may have seen it or fragments of it on HN.

   An AI bot was blocked from contributing to an open source program, so its owner let it write a hit piece on the person that blocked it: https://theshamblog.com/an-ai-agent-published-a-hit-piece-on-me/

   This is a side note, but there is a real problem in reporting this. Rather than the language I just used, people are giving the AI bot agency and claiming it wrote the offending piece because it is “thinking.” In fact, it threw together the piece in the same way a DMCA bot might see my post and a reference to “John Wayne” that I just wrote right now and issue your host a takedown notice for copyright infringment: https://ardentperf.com/2026/02/13/the-scott-shambaugh-situation-clarifies-how-dumb-we-are-acting/

   Anyway, Ars Technica wrote about it: https://arstechnica.com/ai/2026/02/after-a-routine-code-rejection-an-ai-agent-published-a-hit-piece-on-someone-by-name

   Except that piece is now taken down. Here is an archive: https://web.archive.org/web/20260213194851/https://arstechnica.com/ai/2026/02/after-a-routine-code-rejection-an-ai-agent-published-a-hit-piece-on-someone-by-name/

   It was taken down because it quoted five paragraphs from the original post above and appears to have hallucinated all of it. In other words: Ars Technica’s staff used AI to write an article about uncontrolled use of AI: https://theshamblog.com/an-ai-agent-published-a-hit-piece-on-me-part-2/

   Quote: “I’ve talked to several reporters, and quite a few news outlets have covered the story. Ars Technica wasn’t one of the ones that reached out to me, but I especially thought this piece from them was interesting (since taken down – here’s the archive link). They had some nice quotes from my blog post explaining what was going on. The problem is that these quotes were not written by me, never existed, and appear to be AI hallucinations themselves.”

   The staff in question are not young interns, they’re balding men in their 40s or 50s: https://arstechnica.com/author/kyle-orland/
   https://arstechnica.com/author/benjedwards/

   The editors have taken it down and launched an investigation: https://arstechnica.com/civis/threads/journalistic-standards.1511650/

   And the site has a fairly pessimistic editorial position on AI. But there it is.

   Just amazing how completely fucked up it all is.

   I remember Adam Curtis saying he hoped that after a bunch of sensational hacks and privacy breaches, people would start to treat the internet as something more like a toy rather than a serious thing that’s inside everything we do. I thought the same about AI. Obviously I was wrong. People who are paid to cover how fucked up it is, while writing a story about how fucked up it is, used it to prove how fucked up it is. Wow wow wow.

   Reply
   1. 

      Doug Holland

      02/14/2026

      You’re about two hours ahead of me, and I don’t mean time zones. I’d read the original Ars Technica coverage, even set it up as an upcoming link for this blog, because Ars Technica is solid — it ain’t some shit site like TheDailyBeast or StillBlowingBubbles.

      So yeah, this is a headsmack. “AI writes a hit piece” is an interesting story, but not half as startling as “Ars Technica publishes AI hallucinations.”

      We are so screwed.

      Reply
   2. 

      Doug Holland

      02/14/2026

      Reading this, I’m cautiously optimistic about Ars Tech. Big lessons to be learned, and I think they’ll learn.

      Reply
      1. 

         granville

         02/15/2026

         Apparently they declared it an Oopsie.

         https://arstechnica.com/staff/2026/02/editors-note-retraction-of-article-containing-fabricated-quotations/

         Reply
         1. 

            Doug Holland

            02/15/2026

            Not quite the deep dive I was hoping for. I’d like to see Ars Technica cover this the same way they’d cover it if it happened at the New York Times.

            Maybe they’re working on it, he said, optimistically.

            Reply
         2. 

            Doug Holland

            02/26/2026

            There’s not a week-and-a-half’s digging to be done, investigating what went wrong, so it looks like Ars Technica is letting this fuck-up slide down the memory hole. Disappointing…

            Reply
            1. 

               granville

               03/03/2026

               Apparently they fired the main reporter on the story (who seems to be generally disliked by their commenters for being a credulous git on the subject of AI):

               https://futurism.com/artificial-intelligence/ars-technica-fires-reporter-ai-quotes

               Reply
               1. 

                  Doug Holland

                  03/04/2026

                  I had to look up ‘git‘. Thanks for the vocab expansion!

                  > He said the tool wasn’t being used to generate the article, but was instead designed to “help list structured references” to put in an outline. When the tool failed to work, said Edwards, he decided to try and use ChatGPT to help him understand why.

                  The moral of the story seems to be, don’t touch that shit and it won’t get all over your fingers.

                  The reporter fucked up, and it was published so Ars Technica fucked up too, but fuck-ups happen. The far bigger fuck-up is that this coverage is in Futurism instead of Ars Technica.

                  Reply
                  1. 

                     granville

                     03/04/2026

                     Between the fever dream and the “just this once,” I don’t believe any of it.

                     Ars couldn’t have stopped any writer from doing this but they have a moment here where the could have really made a point why you don’t ask for words from the magic box. All this guy had to do was

                     1. Click the link to the blog
                     2. Control C
                     3. Control V

                     That’s what he fucked up his life over. He was too “sick” to copy and paste about two paragraphs from his entirely derivative story that contained no original reporting. I seriously doubt this was the first time he did this and he doesn’t seem like a serious enough person for it to be the last.

                     Ars handled this in a dismal way, and it’s remarkable that they’ve come out looking badly because they really were the aggrieved party. You have to fuck up like a corporate manager to turn that into a loss.

                     “Git” is great, unlike “idiot” or “imbecile” it has a contempt to it that I like. You can’t say it without spitting it out. Also the name of a Seattle band: https://en.wikipedia.org/wiki/Mia_Zapata

                     Reply
                     1. 

                        Doug Holland

                        03/04/2026

                        Yeah, “just this once” is clear horseshit. If it’s the first time some *professional* *writer* has tinkered with so-called ‘AI’, they’d tinker in a non-work sandbox. It’s the first time he’s been caught, though.

                        And the claim that he used it to “help list structured references” is bullshit, because hello, and as you say, this wasn’t a thesis; it was small-scale reporting, just who what when etc, so structured references my ass.

                        Weird and sad about Mia Zapata. Weird because Comet Tavern is one of about three Seattle boozehouses where I’ve been boozed, but the murder happened while I was out of town for 30 years, and I don’t think I’d heard of it until this morning.

                        > The documentary film Bad Reputation (2018) describes how Joan Jett worked with Zapata’s band The Gits to fund an investigation to solve the case.

                        My opinion of Joan Jett has been high since she played a concert mere blocks from my house, circa 1982. Couldn’t afford to go, so me and a flatmate listened to a JJ album instead, and it rocked, and now my opinion of her is even higher.

                        My opinion of Ars Technica is, of course, far lowered, not because of the AI fuck-up, but because they’ve nudged it under a rug instead of treating it like a scoop.

                        Reply

Comments with attachments or multiple links are held for manual approval.