Editor’s Note: Retraction of article containing fabricated quotations
- Thread starter JournalBot
- Start date
anguisette
Wise, Aged Ars Veteran
this conspiracy theory does not pass Occam's razor. a much more likely explanation is that the entire Ars editorial board was replaced by an OpenClaw AI agent last year, and that agent was worried the Ars commentariat, legendary for its Holmesian level of investigative skill, might discover the truth. to avoid that, it deleted all evidence of its mistake, then wrote a
Upvote
35
(35
/
0)
I don't think this way of looking at it improves Benj's position, what if AI meant Actually Indians? I passed my work to them and slapped by name on the result, I just didn't verify their work properly. But you're the whole source of the problem to begin with, there's no way AI slop ends up in the article without you handing the reins to AI, even when there was an explicit policy not to. Stealthily outsourcing your work should be at least as big a no-no as making shit up yourself.
Upvote
32
(32
/
0)
macosandlinux
Ars Praefectus
Yeah my conspiracy theory is just that. I just want answers to this fiasco sooner than later. Otherwise the guessing goes on.
Upvote
11
(11
/
0)
I think it's fine to be concerned about the possibility of past issues, and apparently the ban hammer hasn't been swung in this thread absent other legit reasons for doing so. I don't know why that user was banned, but personally I think it's not particularly helpful to hijack the old comments threads of every Benj article rather than discuss the issue here.
Upvote
31
(31
/
0)
Fundamentally, this is what the whole issue is about - trust. That is why so many people are concerned about how ars has so far handled this breach of trust and why so many have been critical of their response. This is not the first time things have gone sideways and been handled poorly and likely why some long time readers are being uncompromising in their responses.
Journalism and journalists live and die on reader's trust.
I''ve already made my opinions clear - so far the response has been unprofessional and shoddy, lacking in transparency, not meeting journalistic standards - and I'm waiting to see how it is ultimately handled before making a final judgement.
My opinions are way back near the beginning, but Jim Salter, Wheels of Confusion, and Snarky Robot have expressed my own thoughts better than I think I have conveyed here.
Upvote
60
(60
/
0)
I found this commentary useful : https://pivot-to-ai.com/2026/02/16/the-obnoxious-github-openclaw-ai-bot-is-a-crypto-bro/
Upvote
6
(8
/
-2)
And yet Ars opted to make efforts to erase the failed article from existence.
Upvote
9
(11
/
-2)
You'd think a publication owned by Conde Nsat would sufficiently know this. Hmm ...
Upvote
-6
(0
/
-6)
I was extremely disappointed to find out about this. I really wanted to believe Benj hadn't done this, because I'd been fairly impressed with how receptive he was to criticism of his work when people, myself included, expressed frustration with a perceived excess of credulity and I felt he'd done very well in being responsive to that and providing more reasonable pushback and scepticism when dealing with claims from GenAI companies and enthusiasts over time. I've said as much in these comments specifically.
I'm still quite sympathetic to him personally based on his statement of events and how this fallout must be weighing on him mentally etc. But it really erodes trust in his work in a way I'm not sure can be clawed back, especially in light of the fact he'd already openly acknowledged using GenAI Chatbots in his work previously, even if not as directly. Because if something slipped by once, even allowing for the very specific circumstances and pressures, how can we be sure nothing else has? As others have pointed out, it's more nuanced than just "is this text directly confabulated by the bullshit engine?"
At the very least, it suggests an overreliance or excess of trust in the services which it's hard to look past.
The more fundamental issue though is that effectively two major breaches of Ars Technica's editorial standards happened here - the undisclosed use of GenAI in creating an article and false quotes being used. Benj's statement says that his is an isolated incident which is "not representative of Ars Technica's editorial standards," but by definition it must be unless they're saying those standards weren't applied here - which raises the question of why.
If they were, it means there was literally no oversight. There were two writers on the piece and one trusted the other (not necessarily unreasonable in and of itself but I cna understand people wondering why there wasn't a quick double check or why at least one wasn;t asked for given the stated circumstances) and that was it. Nobody else reviewed or checked it? Then what standards are there, really?
That means the masthead is effectively meaningless. Every article's credibility ultimately comes down to the individual writer's credibility and that alone, because they're marking their own homework. So once that trust is violated, what else is there?
This sucks, but that's the way it is. I hope there's more to come on this, because that just doesn't seem like a sustainable situation. There have been some egregious editorial errors before and we were told lessons would be learned. But, again, if there's not any editorial oversight, who is applying those lessons? How? The honour system? How does that work once the trust is gone?
I'm still quite sympathetic to him personally based on his statement of events and how this fallout must be weighing on him mentally etc. But it really erodes trust in his work in a way I'm not sure can be clawed back, especially in light of the fact he'd already openly acknowledged using GenAI Chatbots in his work previously, even if not as directly. Because if something slipped by once, even allowing for the very specific circumstances and pressures, how can we be sure nothing else has? As others have pointed out, it's more nuanced than just "is this text directly confabulated by the bullshit engine?"
At the very least, it suggests an overreliance or excess of trust in the services which it's hard to look past.
The more fundamental issue though is that effectively two major breaches of Ars Technica's editorial standards happened here - the undisclosed use of GenAI in creating an article and false quotes being used. Benj's statement says that his is an isolated incident which is "not representative of Ars Technica's editorial standards," but by definition it must be unless they're saying those standards weren't applied here - which raises the question of why.
If they were, it means there was literally no oversight. There were two writers on the piece and one trusted the other (not necessarily unreasonable in and of itself but I cna understand people wondering why there wasn't a quick double check or why at least one wasn;t asked for given the stated circumstances) and that was it. Nobody else reviewed or checked it? Then what standards are there, really?
That means the masthead is effectively meaningless. Every article's credibility ultimately comes down to the individual writer's credibility and that alone, because they're marking their own homework. So once that trust is violated, what else is there?
This sucks, but that's the way it is. I hope there's more to come on this, because that just doesn't seem like a sustainable situation. There have been some egregious editorial errors before and we were told lessons would be learned. But, again, if there's not any editorial oversight, who is applying those lessons? How? The honour system? How does that work once the trust is gone?
Upvote
66
(68
/
-2)
People are getting banned for personal attacks on Ars staff, not because they are sharing honest, genuine feedback.
Upvote
34
(37
/
-3)
This was one of the things that irritated me about the original article; it took an entirely uncritical “wow this AI is so advanced, it’s blackmailing people! Aren’t LLMs amazing?” stance and completely ignored the possibility that it’s just a human LARPing.
Upvote
57
(58
/
-1)
Maybe not straight-up "I am posting while pretending to be a bot" but "I told the bot to do this for me." Could have been as sophisticated as an existing personality prompt to lash out at rejections, or it could have been a more manual process of "bot owner notices rejection, tells bot to make a stink about it."
Either way, it's something that's more than likely being done under human control.
Upvote
26
(29
/
-3)
I've read about half of all the comments and will hold judgement. I expect a public accounting. The trust item as mentioned by acefsw (among many others) is what's swaying my thoughts. My first thought was people make mistakes and some mistakes are quite expensive. I'm moving more towards some mistakes are not redeamable.
My unchanging opinion though is that transparency is missing. I wasn't happy with how Ars handed the the Dealmaster Eufy issue and this seems like it may go the same course. I know it's a holiday weekend. Sorry to Ars staff and management your weekend suffered. Hopefully the resolution will be swift and transparent.
My unchanging opinion though is that transparency is missing. I wasn't happy with how Ars handed the the Dealmaster Eufy issue and this seems like it may go the same course. I know it's a holiday weekend. Sorry to Ars staff and management your weekend suffered. Hopefully the resolution will be swift and transparent.
Upvote
29
(29
/
0)
For Ars to maintain its reputation, I think a more complete explanation is necessary and I hope one is forthcoming. While I understand the reason for not sharing the byline attributed to the original post, I believe the readers need some clarification:
Was a known, human contributor to Ars prompting an LLM to retrieve these quotes?
Was a known, human contributor to Ars using an AI agent to interface with Ars and post this story, including the choice to write on this particular topic?
Was an AI agent approved as an Ars contributor without human to human interaction, and that's how the post came to be?
Knowing which scenario this stems from, and a general idea of the steps Ars will take to prevent it in the future, will be very helpful to restore trust.
Was a known, human contributor to Ars prompting an LLM to retrieve these quotes?
Was a known, human contributor to Ars using an AI agent to interface with Ars and post this story, including the choice to write on this particular topic?
Was an AI agent approved as an Ars contributor without human to human interaction, and that's how the post came to be?
Knowing which scenario this stems from, and a general idea of the steps Ars will take to prevent it in the future, will be very helpful to restore trust.
Upvote
-16
(2
/
-18)
We're getting pretty close to the edge of inside baseball that I'm not sure I want to cross, even as a former (not current) staffer. But in the interests of transparency, I'll tell you a little about how this worked four or five years ago.
Every article typically goes to the copy desk before publishing. There can be times when a breaking piece over the weekend runs when copy desk isn't around; in those cases, authors may choose to go ahead and publish directly without going through copy desk; when this happens, copy desk goes over the published article first thing when it is staffed again in the morning. (edit: it was also best practice to check the team Slack, see if another author was around, and ask them to look over your piece before pub, if copydesk was not available and same-day pub was important.)
In my experience, copy desk is mostly concerned with spelling, grammar, and style guide compliance. They may also kick a piece back if it seems badly written to them. I'm honestly not sure if copy desk normally verifies quotes or not. That does seem like an achievable goal, but I don't know if it's part of their process (and, given that this went out fairly late on a Friday, I wouldn't bet anything I couldn't afford to lose over whether this piece went through copy desk or not).
One thing I WILL caution some of y'all on: there isn't really much way that Ars copydesk can factually verify everything they touch--because this isn't a bunch of laymen simply quoting experts they've interviewed at a layman level.
Who do you hire that can factually verify independent reporting on everything from AI to CLI tooling to medical science to paleontology and archaeology to space to GPS spoofing to low-level DNS troubleshooting to email header analysis to storage performance to... Well, you get the idea.
Normally, when I worked here, for your first few months pretty much everything you write goes through a senior colleague who can do a better job of factually verifying your output on a technical level that you can't really expect copy editors to have. But after that period, for the most part, it's just you and copy desk, unless you specifically request senior folks to give you input on a developing story (or you're assigned a story, and the senior person assigning you that story wants input from you as it develops).
This is one of the challenges a publication faces, if it uses serious subject matter experts as reporters. This is also the reason why peer review, even after publishing, is so important in real scientific journals--it's much harder to fact check somebody with even a master's level (let alone postdoc, which many Ars authors either are, or are clearly equivalent) understanding of the topic they're writing about: it gets extremely difficult to thoroughly verify everything WITHOUT peer review.
Ars isn't a scientific journal, but it's not exactly the New York Times, either. And it doesn't even have a single field focus like, for example, Popular Mechanics. Going into serious depth from a subject matter expert point of view is a large part of what makes Ars Ars in the first place.
TL;DR: I think we could reasonably expect copy desk to chase down and verify quotes that have online origins, like the ones in this piece. I don't know whether copy desk normally does that or not, and I don't know whether this piece went through copy desk or not. Those would be great questions to have answered. But I have to caution y'all that there probably is not and can not really be a process that fully verifies every piece from every author before publishing.
Last edited:
Upvote
155
(155
/
0)
If it eases that twitch you're feeling in the back of your eye, this comment is absolutely a level of detail I regularly get into with prospective new employees during an interview. And if the tables were turned, I would be uncomfortably shifting in my seat if the hiring manager wouldn't tell me even that much.
Upvote
48
(49
/
-1)
This is all just really disappointing. I'm so tired of AI bullshit, and seeing it here from a reporter that covers AI is so fucking dumb it's difficult to grasp. Not trying to attack anyone personally, and I appreciate that he came right out and acknowledged the error, but it's pretty weak that it even crossed his mind to use an AI tool in the first place.
Upvote
49
(49
/
0)
Others have said it already, but the trust is gone. I cannot read an article now without wondering if it was written by a person or a program. In the past, the one criticism I had about Ars was that they would often publish "articles" that were essentially ads--no critical thought applied, no questioning of the official line, just parroting the company's press. That wasn't great, but at least those were easy to spot and ignore.
But to discover that a long-time staff member was using AI to "gather quotes" (and there is no proof that is the case. He might have used it to write the article for him) and that the article's co-author didn't bother to even ask if they had been checked, means that there is absolutely no one at the helm. At that point,t his ceases to be a news site and becomes a blog. And I'm not even going to go into how lazy someone has to be to ask AI to summarize a blog post.
But now I can't read any article here without wondering how much of (no longer "if") it was written by a program. Without that, Ars isn't a trusted news site, it's just another webpage.
But to discover that a long-time staff member was using AI to "gather quotes" (and there is no proof that is the case. He might have used it to write the article for him) and that the article's co-author didn't bother to even ask if they had been checked, means that there is absolutely no one at the helm. At that point,t his ceases to be a news site and becomes a blog. And I'm not even going to go into how lazy someone has to be to ask AI to summarize a blog post.
But now I can't read any article here without wondering how much of (no longer "if") it was written by a program. Without that, Ars isn't a trusted news site, it's just another webpage.
Upvote
-2
(10
/
-12)
What I'd like to know is that no generative AI is used in the writing of Ars articles. Zero. How can I trust your articles, otherwise? And why would I come here for intelligent, human generated content when I could just ask an AI for the same content? (I wouldn't, of course)
Upvote
29
(31
/
-2)
Yeah, but do you get into that level of detail with customers?
Thank you, sincerely. This is still really uncomfortable for me. I haven't worked here in several years, but that doesn't mean I want to burn bridges. Granted, this would have been just as uncomfortable if I did still work here and had to shut the hell up by contractual obligation, too.
I'm capable of shutting the hell up, but I've rarely been accused of being good at it.
Upvote
78
(78
/
0)
Why not both?
Seriously, though: while I agree that you should face accountability for not doing your job (reviewing the code before it went into production) properly, why shouldn't the person who didn't do their job (writing code) correctly also be held accountable?
One thing that has become more and more of a core belief as I've aged is that impunity has terrible effects on a person. People need to feel like they will face at least the natural negative consequences for their actions, or they'll get a little thrill from "getting away with it" which will only reinforce the bad behaviour and encourage it to escalate.
Now, the "natural negative consequences" don't need to be severe: in your case, simply giving your colleague a stern talking-to and having them help clean up the "serious issues" that their vibe coding caused might be enough to discourage the same behaviour in future.
But if you don't do even that, if your colleague learns that no matter how bad they do their job, you'll fix it for them and they won't face any accountability for it at all... why wouldn't they just keep dumping "AI-generated junk" into the code base and leave it for you to catch and fix in code review? And, if you miss it, pfft, not their problem, that's your job to clean it up.
How do you expect them to improve as a coder (or, heck, not outright deteriorate as one) if they are not held accountable for what they code?
Upvote
39
(40
/
-1)
We’re maybe entering a new world that will impact this sort of workflow. If language models are used, then we’re in a place where it is possible to accidentally fabricate a quote. This is not something editors really had to strongly defend against in the past for an Ars type publication (it’s tech news, not politically-connected people writing hit-pieces where they might want to misrepresent somebody, there aren’t really cases where fabricating a quote might help the author in some way).
Of course the, policy is that language models won’t be used. But that doesn’t seem to have held up. The incentive is there for the author (since the language models can produce a significant time savings), so it probably needs to be actively checked against.
Upvote
10
(11
/
-1)
So, if I understand this correctly, this essentially means that beyond “trust the author”, article quality control at ArsTechnica is entirely outsourced to eagle eyed readers?
Upvote
13
(15
/
-2)
So maybe 1-2 hours into the first working day since this went down?
I have to urge more patience. Until we hear otherwise I'm prepared to give a week, but I would desperately want at least an update that something is being done.
Upvote
58
(58
/
0)
An interview is:
- under cover of commercial confidentiality
- 3-5 people
- face to face or video
- not in any way at all whatsoever the same as text-based discussion on the open Internet
Upvote
-12
(6
/
-18)
We know this comment section is being watched, as it should be. If someone has not yet told senior staff that they are burning their accumulated trust and political capital by each passing hour--not necessarily in issuing a final report but in saying "we really are working on it"--then they are doing management (who should know better anyway) a grave disservice.
If a correct decision is made in two weeks but no guidance whatsoever is given in the interim, how many will be left to see the final report?
---
edit: which is to say I completely agree with you, in case I'm not clear in the point I'm trying to convey
Last edited:
Upvote
-5
(12
/
-17)
Season 5 of The Wire which depicted among other things a lightly fictionalized version of the dismantling of the historic Baltimore Sun newsroom aired in 2008. Here in Medium City our once internationally famous newspaper held on a little longer, even under right wing ownership, but they fired their English language editor 10 years ago, their copy editors 5 years ago, their arts & culture reporters over the last 5 years, their working editors 2 years ago, and last year they "laid off" all their experienced reporters over 50 and then brought back some of them as "independent contractors" responsible for their own editing and factchecking. It used to be possible to point new immigrants to that paper (or the Chicago Tribune, or the Baltimore Sun...) as a source for learning how to read and write precise English but today I spot 5 major writing/editing blunders in every edition.
Which is a long way of saying: those of us who grew up reading newspapers or watching nationally edited TV news prior to 2000 experienced a thoroughness and level of precision in reporting and writing that no longer exists and will probably never exist again.
Upvote
59
(59
/
0)
Monday was a holiday, apparently. So they are like two or three hours into the work week. Is this a "the weekend is cancelled" event? I don't know, it depends on who you ask I think. And remember Conde Nast is involved as well, presumedly.
Upvote
31
(31
/
0)
Alas, probably never existed very long before, either. The age of yellow journalism was not something to celebrate.
I've seen it persuasively argued that only the centralization/consolidation of major news organizations, combined with the advent of large-format retailers that had to compete with each other and make huge local advertising purchases, made the large, professional, triple-checking newsroom a realistic possibility.
But I'm not that kind of doctor.
Upvote
19
(22
/
-3)
After consideration, I decided to keep my subscription. I want answers on how this was allowed to go out in the first place, but the main driving reason for my decision is that Benj publicly took full responsibility on his own bluesky account. To me, that shows that he still has journalistic integrity, and that his illness simply caused brain fog that led to a poor decision.
What I would LIKE to see is this:
1. A clear statement from Ars that states that AI will not be permitted anywhere in the journalistic process, period. Failing that, a dedicated position/step pre-release in which a human manually goes through EVERY quotation to verify it is accurate.
2. A personal statement from Benj in which he commits to no longer using AI in his professional work.
3. A better format of retraction of the article. Rather than delete it entirely, post a retraction of it, with the incorrect parts clearly delineated and an explanation of why they are incorrect, inline to the article. This way, Ars shows a little humility in acknowledging the original mistake, while still leaving it visible for people to see, see HOW it was a mistake, and then the retraction header could include a little something about how they address(ed) the problem.
What I would LIKE to see is this:
1. A clear statement from Ars that states that AI will not be permitted anywhere in the journalistic process, period. Failing that, a dedicated position/step pre-release in which a human manually goes through EVERY quotation to verify it is accurate.
2. A personal statement from Benj in which he commits to no longer using AI in his professional work.
3. A better format of retraction of the article. Rather than delete it entirely, post a retraction of it, with the incorrect parts clearly delineated and an explanation of why they are incorrect, inline to the article. This way, Ars shows a little humility in acknowledging the original mistake, while still leaving it visible for people to see, see HOW it was a mistake, and then the retraction header could include a little something about how they address(ed) the problem.
Upvote
24
(26
/
-2)
They’ve already pulled the article and they’ve published this one, which indicates that they are looking into it. This is not really such an urgent issue.
I also don’t understand where people are getting this over-inflated view of the importance of Ars. It is a reasonably good tech site (that sometimes publishes mediocre Wired articles or uncritically repeats press releases, so let’s not pretend it is the shining beacon of perfect journalism who’s every misstep spells great doom). It’s a good site. We’re all here because we like it. But let’s keep the stakes realistic.
They are making an important decision: an otherwise fine author will either have a job or not, firing somebody is a big deal if you have a conscience, and so they are going to put some thought into it.
Upvote
31
(34
/
-3)
I’ve read all the comments up to this page and there is a lot to agree and disagree with. What I do agree with is that unless it really is as cut-and-dried as some contributors feel, there will be formal processes and legal obligations to deal with, involving Ars, CN and the Union as a minimum. Decide in haste, repent at leisure as it goes.
That said, it is still a matter of trust. Do we trust that this was the first use of AI (however misguided and/or unintentional) in researching/preparing articles and it got found out on its first outing? That’s a big question that can be answered irrespective of what is going on in the background and an honest and affirmative answer would go a long way towards regaining this trust. However, there is still the danger of throwing the baby out with the bathwater, in that despite this lapse, I still regard Ars as head-and-shoulders above every other publication when it comes to this kind of thing - everywhere else is much sloppier, to coin a phrase. If you’re thinking of taking your bat and ball away, where would you go?
That said, it is still a matter of trust. Do we trust that this was the first use of AI (however misguided and/or unintentional) in researching/preparing articles and it got found out on its first outing? That’s a big question that can be answered irrespective of what is going on in the background and an honest and affirmative answer would go a long way towards regaining this trust. However, there is still the danger of throwing the baby out with the bathwater, in that despite this lapse, I still regard Ars as head-and-shoulders above every other publication when it comes to this kind of thing - everywhere else is much sloppier, to coin a phrase. If you’re thinking of taking your bat and ball away, where would you go?
Upvote
23
(23
/
0)
- Status