Talk:Russian sabotage operations in Europe
| This article is rated C-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects: | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Did you know nomination
[edit]
- ... that European officials have reported a surge of suspected Russian sabotage acts across the continent—including arson, GPS jamming, and infrastructure attacks—since 2022? Source: https://www.economist.com/europe/2024/05/12/russia-is-ramping-up-sabotage-across-europe
- ALT1: ... that Russian sabotage in Europe is a type of hybrid warfare? Source: https://www.iiss.org/online-analysis/online-analysis/2024/11/russias-hybrid-war-in-europe-enters-a-dangerous-new-phase/
- ALT2: ... that a Russian campaign of sabotage in Europe is suspected to be ongoing since 2022?
- Reviewed: Template:Did you know nominations/Harmony Cobel
Piotr Konieczny aka Prokonsul Piotrus| reply here 03:54, 7 September 2025 (UTC).
I think this is biased. I think other users probably think the same. And that's not interesting, because it's news that's widely reported around the world. Cassilvwikis (talk) 02:01, 10 October 2025 (UTC)
An actual review required, preferably one who understands WP:DYKG.--Launchballer 21:10, 11 October 2025 (UTC)
I've started a separate section Talk:Russian sabotage operations in Europe#Suspected LLM usage, because the article still requires a huge amount of editing to remove any hint of the text being written by an LLM. Boud (talk) 12:44, 25 October 2025 (UTC)Just to clarify: this cleanup is needed before a proper DYK review. Boud (talk) 12:47, 25 October 2025 (UTC)
- Lots of cleaning work still needed. @Piotrus: At least two non-existent, unarchived URLs have been found so far. Both seem to correspond to real articles, but if it's an LLM that chooses the URL, then you have to check it; and if it's you that stores the URL for later use a few years later, best that you archive it, and you do have to check it when you publish the article in Wikipedia. You can't assume that the old URL is still valid. (These are generic "you"s.) DYK reviewers are especially required to check sourcing. With 81 sources, that's a lot of checking that reviewers need to do, and finding any dead-unarchived URLs implies that more than just a quick browse is needed. Boud (talk) 18:44, 25 October 2025 (UTC)
- Plus a third dead URL. I'm about 3/4 of the way through the 81 refs in terms of rapid copyediting, but I have not checked all of the roughly 60 URLs in text that I've copyedited so far; I've assumed with good faith that they exist and that the info in the text matches the content. I've checked some. Boud (talk) 18:54, 25 October 2025 (UTC)
- WP:DYKNEW
.WP:DYKLEN
.WP:DYKCOMPLETE
.WP:DYKCITE
Most seem OK, but as above, I've checked quite a few of the sources, but not all, and there were at least twothree dead, unarchived (from a 6 September 2025 new article!) URLs that I've fixed, and there still is one unfixed (handelsblatt).WP:DYKHFC The lead currently has inline refs for arson, GPS jamming and attempted PL railway attacks (infrastructure), but I haven't really tried to check about whetherEuropean officials ... reported
all of these.WP:DYKHOOK I don't see the point ofEuropean officials ... reported
in the original hook. I don't see much point to ALT1 either, since the lead of hybrid warfare statesThe concept of hybrid warfare has been criticized by a number of academics and practitioners, who say that it is vague and has disputed constitutive elements and historical distortions
and in this particular case, Russia is currently only using non-military methods of sabotage in Europe west of the Ukrainian border, so it's semi-hybrid rather than hybrid, which is likely to leave readers even more confused about what that actually means. I would propose ALT2 (above).
Overall, with ALT2, this article is not too far from DYK ready. However, it does need someone other than me to do another review, especially in checking (1) (preferably) all of the references, and checking that they are actually used acccurately; and (2) checking if it's justified to remove the {{AI-generated}} template and category. I found a fourth dead, unarchived URL. It's not credible to believe that the four URLs became invalid in less than two months. Mainstream news source servers sometimes do server software updates and fail to create redirects for old URLs. But that's once in five years or so, not every month. So overall, this needs another, independent reviewer. Boud (talk) 21:52, 25 October 2025 (UTC) - @Boud: Thanks for the c/e. I removed the broken handelsblatt ref which was not necessary, added others and clarified this. As for the dead https://www.telegraph.co.uk/news/2021/04/18/czech-republic-expels-18-russian-diplomats-2014-ammunition-depot hmmm. The correct one is https://www.theguardian.com/world/2021/apr/19/russia-expels-20-czech-diplomats-in-tit-for-tat-response and/or https://www.theguardian.com/world/2021/apr/18/dominic-raab-uk-fully-supports-czech-hunt-for-skripal-suspects . Not sure how that ref went dead, or if it was malformatted during the AI c/e. I thought I checked and verified all refs in that article, but just in case, I'll be extra carefull to double check if all refs work after I use such tools. --Piotr Konieczny aka Prokonsul Piotrus| reply here 04:33, 26 October 2025 (UTC)
- WP:DYKNEW
- Plus a third dead URL. I'm about 3/4 of the way through the 81 refs in terms of rapid copyediting, but I have not checked all of the roughly 60 URLs in text that I've copyedited so far; I've assumed with good faith that they exist and that the info in the text matches the content. I've checked some. Boud (talk) 18:54, 25 October 2025 (UTC)
- Lots of cleaning work still needed. @Piotrus: At least two non-existent, unarchived URLs have been found so far. Both seem to correspond to real articles, but if it's an LLM that chooses the URL, then you have to check it; and if it's you that stores the URL for later use a few years later, best that you archive it, and you do have to check it when you publish the article in Wikipedia. You can't assume that the old URL is still valid. (These are generic "you"s.) DYK reviewers are especially required to check sourcing. With 81 sources, that's a lot of checking that reviewers need to do, and finding any dead-unarchived URLs implies that more than just a quick browse is needed. Boud (talk) 18:44, 25 October 2025 (UTC)
Suspected LLM usage
[edit]This article was created by an extremely experienced Wikipedian with text that violates WP:RELTIME, MOS:", MOS:', and generally uses unencyclopedic tone: assessed that the Kremlin
, Although authorities generally stopped short of directly blaming Russia in those cases, the incidents fit a broader context
, revealed that ... reportedly working at the behest of
, often causing fear and chaos
, orchestrated by
, Poland's prime minister announced that evidence
, campaign's hallmarks
, Nonetheless
, Cruc[i]ally
, analysts believe
, However, some
, officials caution
, tried to deter Moscow
, heightened concerns
. The actual content and references, based on a quick, incomplete browse, look like they're correct and valid. My inference is that an LLM was used to write a summary of a given (valid) list of sources, and there was some double-checking prior to adding the content to Wikipedia. This still requires significant cleanup. I've added Category:Articles containing suspected AI-generated texts to the article; a bot should add AI-generated
automatically.
@Piotrus: If you did actually use an LLM for summarising, I would strongly recommend that you follow WP:LLMDISCLOSE, even though it's not (yet) a formal policy: Every edit that incorporates LLM output should be marked as LLM-assisted by identifying the name and, if possible, version of the AI in the edit summary. This applies to all namespaces.
If you just happened to mimic LLM style, then sorry! At least we now have an alert to human Wikipedians that the article gives a strong impression of being LLM-based, so that they know that a thorough cleanup WP:RELTIME + WP:TONE + MOS:" + MOS:' is needed. Boud (talk) 12:40, 25 October 2025 (UTC)
- @Boud Thanks. I occasionally use LLMs to copyedit my language, as I am not native speaker and I treat AIs as an advanced spellchecker. And I find its ability to summarize content useful (with the caveat that it needs to be double checked, of course, since hallucinations can creep into anything that AI does). Thanks for letting me know about LLMDISCLOSE best practice, I'll add a talk page notice or edit summary in the future. Could you elaborate on what is wrong with the tone of the examples you quoted below? I can't say I see any tone problems there (I do proofread text after the AI does its copyedit pass, and the fragments above did not raise any flags for me). Piotr Konieczny aka Prokonsul Piotrus| reply here 12:47, 25 October 2025 (UTC)
- Hold that thought, I did not notice s WP:RELTIME, MOS:", MOS:' links, I'll review them ASAP. Piotr Konieczny aka Prokonsul Piotrus| reply here 12:47, 25 October 2025 (UTC)
- Ok, I'd appreciate more feedback on the MOS issues. I double checked the linked policies and I don't see any weasel / imprecise words that are mentioned at REALTIME in the cited excerpts. As for quotation marks and apostrophes, I frankly never understood relevant policies perfectly, but I also don't recall any problems related to this being raised in regards of my writing. Perhaps AI changed my style in a way that's not good, but in all honestly I am not seeing the problem (yet). Again, I very much would like to learn what is the issue, since I thought AI will help me create a superior prose to what I usually write (I gave up on FA level articles years ago b/c of complains about my prose; which is why I only have been doing GAs in the last decade...). Piotr Konieczny aka Prokonsul Piotrus| reply here 12:53, 25 October 2025 (UTC)
- Glad you've stated that straight out. :) The tone problems sort of overlap with writing-from-a-POV problems: the sources mostly write things from "our" (Europeans') POV. I can only speculate if the "our" POV is based on the LLM getting that from the sources or from your prompts, or both. To help illustrate the problems, here's a list for your original version (some are probably already cleaned up):
assessed that the Kremlin
, - referring to a state or government by its capital city or notable governmental building is newsy style, not encyclopedic style;Although authorities generally stopped short of directly blaming Russia in those cases, the incidents fit a broader context
, - if they didn't blame Russia, then either we say nothing, or we say that they "hinted" that Russia was to blame, but then that risks requiring Wikipedians to read between the lines of the source; secondly: here we state in Wikivoice that the incidents are necessarily part of a broader pattern, which risks WP:OR, at least internally within this sentence;revealed that ... reportedly working at the behest of
, - "revealed" tends to imply that we know what the truth is necessarily, and the source finally informed us about this truth; "behest of" is rather idiosyncratic style;often causing fear and chaos
, - this sounds quite tabloidy - surely the reader can decide for him/herself what the effect of arson and bomb hoaxes are on societyorchestrated by
, - arbitrary metaphorPoland's prime minister announced that evidence
, - unless the PM is suspected of lying (or just saying whatever sounds good for a mafia-style deal à la Trump), we can consider him/her/them as a reasonable source for that type of statement and just say "evidence"campaign's hallmarks
, - arbitrary styleNonetheless
, - tends to tell the reader which statements are more credibleCruc[i]ally
, - tells the reader which statements are the most importantanalysts believe
, - "analysts" is WP:WEASELlyHowever, some
, - tends to tell the reader which statements are more credibleofficials caution
, - what counts as a warning is a POVtried to deter Moscow
, - citification of a stateheightened concerns
- who has the concerns? "us"? from whose POV are these "concerns"?
- Since you've asked for help, I guess I and others will have to do some copyediting. I'll see if I can have a go fairly soon ...
- Boud (talk) 13:16, 25 October 2025 (UTC)
- @Boud Thanks for taking the time to provide detailed feedback, it is very much appreciated. To address these:
- re: Kremlin - I thought an occasional use of that style would be good for Wikipedia's readability (as you may known, in academic assessments of Wikipedia quality, we score above most other encyclopedias in all metrics but one: prose/readability...)
- although authorities - good catch. Having re-read this, it says nothing, and does sound LLMish. Deleting now.
- reveal - thanks, I see this is at WTW. Will try to remember
- fear and chaos - unnecessary indeed, deleted
- orchestrated - this one is not in the WTW and while I could simplify it to organized, I do like to "spice up" the prose a bit (again, per academic comments about our prose problems)
- "Poland's prime minister announced that evidence" - I think that was c/e-ed already. But isn't attribution usually a best practice? Not that I dispute that shortening it is fine.
- "campaign's hallmarks, - arbitrary style" - could you elaborate on this one? To me it sounds neutral and "eloquent"
- "Nonetheless" - I removed it as likely unncecessary, although I am fairly sure this is unlikely to be AI wording - I think it's a word I use a lot. I don't see it at WTA...?
- "Cruc[i]ally" - ditto
- "analysts believe" - fair, shortened/rewritten
- "However" - huh, fair, and it is at WTW. Removed.
- "caution" - this is not at WTW, but I removed it (shortened the sentence).
- Moscow - ditto... I am not convinced this is bad style?
- "heightened concerns" - fair. C/e done.
- Piotr Konieczny aka Prokonsul Piotrus| reply here 13:35, 25 October 2025 (UTC)
- Re:
prose/readability
+sounds neutral and "eloquent"
. I don't expect the concerns about prose/readability to be a complaint that we don't use enough academic style pontification (WP:BUZZ:and in contexts beyond the business world such as academia
). Writing accurately, satisfying all Wikipedia policies and guidelines, can often lead to writing that is accurate and sourced, but sounds awkward. This is unfortunate, but until someone comes along and copyedits, better to have accurate content than smoothly written content. Advertising/propaganda aims to sound good and convince. We don't aim to convince anyone. Stochasticity and intensity in editor activity also affects the language. I think that if there's a talk page debate about any particular sentence in a Wikipedia article, the consensus will generally evolve towards plain English rather than any particular idiosyncratic style. That's a ... hallmark ;) of Wikipedia consensus IMHO.Anyway, this article is big. Back to editing ... Boud (talk) 13:59, 25 October 2025 (UTC)- Several of the references have
author2=Reuters
or similar; some are subscription-only but almost-open-access (requiring Alphabet Inc captcha, unfortunately) when archived; and at least one is a dead, unarchived url. I'm fixing the ones that I find... Boud (talk) 15:25, 25 October 2025 (UTC)- Thanks again. I have removed the ref you tagged as dead, and added more + c/e to clarify the sentence it was appended to. Piotr Konieczny aka Prokonsul Piotrus| reply here 04:36, 26 October 2025 (UTC)
- Several of the references have
- Re:
- @Boud Thanks for taking the time to provide detailed feedback, it is very much appreciated. To address these:
- Glad you've stated that straight out. :) The tone problems sort of overlap with writing-from-a-POV problems: the sources mostly write things from "our" (Europeans') POV. I can only speculate if the "our" POV is based on the LLM getting that from the sources or from your prompts, or both. To help illustrate the problems, here's a list for your original version (some are probably already cleaned up):
- Ok, I'd appreciate more feedback on the MOS issues. I double checked the linked policies and I don't see any weasel / imprecise words that are mentioned at REALTIME in the cited excerpts. As for quotation marks and apostrophes, I frankly never understood relevant policies perfectly, but I also don't recall any problems related to this being raised in regards of my writing. Perhaps AI changed my style in a way that's not good, but in all honestly I am not seeing the problem (yet). Again, I very much would like to learn what is the issue, since I thought AI will help me create a superior prose to what I usually write (I gave up on FA level articles years ago b/c of complains about my prose; which is why I only have been doing GAs in the last decade...). Piotr Konieczny aka Prokonsul Piotrus| reply here 12:53, 25 October 2025 (UTC)




