For the past few hours, the whole world, or at least my LinkedIn feed, has been talking about one thing: Anthropic signed the EU AI Act’...
For further actions, you may consider blocking this person and/or reporting abuse
Thanks for informing us Sylwia!
To be fair, this is expected and not surprised. It has become an issue and I have receive complaints from other people that they despised AI Slop on this platform (regardless of the moderation I have done).
The best things we can do, is the thing we need to for ourselves, which is start to critical think. Our attention span is so awful to the point where everyone is starting to lose this ability to think for ourselves instead of just giving up and following the herd. Yes, AI is everywhere BUT we need to train ourselves to take control and think to ourselves instead of reaching conclusions and I can't stress this enough.
Sorry Sylwia if this sounds frustrating, but it's something I thought about while typing lol. xD
Thanks so much, Francis! And exactly, no watermark is going to save us here. Critical thinking is absolutely key.
Honestly, I don't think it's that bad here on DEV yet, but LinkedIn is a nightmare. 😂 And somehow people still like and share all that AI slop, so apparently... they actually enjoy it? Which might be the most terrifying part of the whole thing. 😅
Oh for sure, LinkedIn is a hive for AI slop. It's already toxic enough for people to compare yourself to other people on that platform and it's a no go. Yea sure, I have a LinkedIn, but I rarely post on there.
It's bad enough where I get reports from @codingwithjiro, @klaudiagrz and other people to the point it's annoying. It's not bad, but it's to a degree where we notice it. Sure, people use it to a good degree for Grammar, but people ruin the fun for that by claiming that, where they didn't. Oh yea, for comments, still have no idea why we see AI-Generated Comments. I can understand posts and all, but for comments is crazy. I understand professionalism, but there are many ways to do this. Was wondering if you knew why this maybe the case?
Yes, AI-generated comments... It's CRAZY!!! IDK, maybe it's a kind of self-promotion? I even understand AI-polished comment, for clarity. But some of them are pure AI slop 😅
Anyway, I'm going to sleep now, I'm not sure what I'm doing here on DEV at 1 AM 😂
OMG, YES! 💯 I understand running a spellcheck over the article to correct typos and so on... but for the comments... come on, I want to hear your voice, your style, your way of expression, not another generic slop 😬 We all communicate in different ways and styles, and that's what's interesting about exchanging opinions on such platforms like DEV.
I can even understand running a spellcheck on a comment. But some comments are literally just generic AI-generated summaries of the article with no original thought added at all. And at that point I don't understand why someone even bothers posting them 😅
Thanks for the writeup. Tbh, I'm glad somebody is finally enforcing this AI content thing. I wrote about it a while back when I got upset with the "ban everything AI" wave that's still hanging around in places. My biggest concern with this is that your distinction
gets collapsed into something resembling the AI version of hellfire and pitchforks...
Great article, nonetheless!
Thanks so much, Ashley! I'm starting to notice this panic too, especially among non-technical people: this idea that anything involving AI is automatically bad.
But people are going to use AI anyway. And just like @francistrdev said, I think we simply need to get better at critical thinking instead of treating everything as either "AI = evil" or "AI = good." 😅
Interesting, thanks for the information, Sylwia. I’m really curious how this will work in practice, especially with code.
At least in our case, we use AI to generate code that has to fit into an existing codebase: naming conventions, patterns, formatting ... So I’m really curious how they want to put a watermark into that and how much of it will survive after the code is adapted, formatted, or refactored.
It looks like I have the wrong friends on LinkedIn though, because my feed definitely didn’t blow up with this. I still mostly see recruiter posts with pictures of them working from the beach. 😂😂
PS: I just had to add this, it perfectly fits the whole EU situation (and it doesn’t even have to be about watermarking). 😂
Hahaha, maybe we'll start getting variables like claudeResponse = await claudeFetchData() just to make sure the watermark survives. 🤣
BTW, I'm genuinely jealous of your recruiters working from the beach! My LinkedIn feed is basically permanent drama, especially Polish LinkedIn. 😂 A very popular genre of post there is basically: "EVERYONE ELSE is publishing AI-generated garbage, but NOT ME. I’m honest, I do everything properly, and I’m better than that."
The beach photos are nice, but all those recruiter mottos and “smart sentences”… I’m honestly getting tired of them, but even so, your feed sounds much more irritating. 😂😅
Oh yes, mine is absolutely terrible. 😂 It's basically: "Everyone on LinkedIn is doing it WRONG! Someone dared to ask a question at the end of their post, probably just for engagement! How dare they! Only I do LinkedIn properly and honestly!" And then underneath: 1,000 likes and hundreds of comments saying, "YES, this annoys me too!"
Meanwhile, LinkedIn itself... if I post something relatively smart, I get maybe 10–20 likes. But if I post a photo of myself making a stupid face? 200+. xDDDDDD
So apparently the algorithm has spoken. 😂
😂😂😂 And if that photo was from the beach, it would probably be 2,000+. 😂😂
That's the next step, once I decide it's time to achieve international fame. 😂😂😂
Ah, Sylwia! Your articles always seem to show up exactly where I need them. I was actually wondering about this watermark earlier tonight.
And what Claude says — and what you point out — pretty much confirms my view: this is a band-aid on a wooden leg.
Written by AI? Written with AI assistance? Corrected by AI? Not AI at all? Or actually written by AI but simply not detected?
In other words, complete nonsense on a scale matching the sheer foolishness of people who think they can know everything about everything and solve every problem with a binary answer.
It reminds me, for instance, of the proposed ban on social media for under-15s, or France’s tax supposedly aimed at blocking Temu, which ultimately did little more than hurt jobs in France — and, incidentally, make Bercy and its brilliant minds look rather ridiculous.
The more complicated the problem, the more absurd it becomes to pretend there’s a simple yes/no solution.
By the way, I wrote this comment in French and had ChatGPT translate it into English. So… how exactly should we classify it? 😉
Exactly, Pascal! And for many people, this is simply a way of leveling the playing field. We're not writing novels here. 😅 Someone translates their text with AI (I sometimes write comments in Polish and translate them too, simply because it's faster), someone else uses it to fix grammar and typos, and so on.
And Anthropic explicitly says that this kind of AI-assisted text can get the watermark too. That's it. So what exactly would detecting that watermark prove?
That's also why I suspect these detection tools won't be universally available to everyone. Otherwise, I can already imagine the absolute paranoia we'd end up with. 😂
I hope so, but unfortunately, I have absolute faith in our institutions: if faced with a choice between an intelligent decision and others that are more questionable, they will invariably choose one of the worst options. They’re basically applying the Dilbert principle...
Hahaha, I'm dying. 😂 I thought we only had decision-makers like that in Poland, but apparently this is a truly global phenomenon.
Hi Sylwia, fantastic breakdown! You perfectly cut through the hype and myths surrounding the EU AI Act and Anthropic's watermark.
Your point about the "Code Watermarking" challenge is the most critical part of this discussion. In natural language, the latent space for synonyms is vast, but in code, syntax is rigid. If we add tools like Prettier, ESLint, or minifiers into the CI/CD pipeline, the statistical bias could easily be flattened out or stripped entirely.
I'm really curious to see Anthropic's technical docs on how they plan to solve the code watermarking problem without breaking linters or refactoring workflows.
Do you have any insight into whether Anthropic has discussed a multi-stage watermarking approach that survives code minification? For instance, minifiers rename variables and remove whitespace—so wouldn't it be more robust to embed the watermark in the AST (Abstract Syntax Tree) structure rather than the raw text? This seems like the real engineering challenge: any watermark that relies on surface-level syntax might be stripped after running Prettier or ESLint.
Thanks for another insightful read! Wishing you continued success and more great technical deep dives!
🗻🌊
Thanks so much for the kind words! 😊 And unfortunately, I can't answer your question about code because... I simply don't know! 😅
Anthropic hasn't published the technical details yet and says the documentation is still coming, so for now, we have to wait. I'm really curious about this part too, especially how (or whether) the watermark could survive things like formatting, refactoring, and minification.
So... Anthropic, we're waiting for those docs! 😄
From the positive perspective, the watermark could help universities to check whether students cheating or violating academic integrity policies by using AI. They may not be the definitive proof, but they could provide an additional signal.
Exactly! Or for detecting deepfakes in general. That's why I have a feeling access to these detectors might be limited to institutions that actually need them, like universities and similar organizations.
In Poland, for example, universities already use an anti-plagiarism system for academic work. And even that can be bypassed relatively easily if someone is good enough at rewriting things. 😅
So I can imagine watermark detection being treated as an additional signal rather than definitive proof. But we'll see how they actually implement access to it. At this point, nothing would surprise me. 😂
I think there is an interesting distinction here between provenance and detection. 🔍 A watermark can be useful as a provenance signal, especially for media like images, audio and video, but I don't think it should be treated as proof that "AI wrote this".
The code example makes this even more interesting. 💻 If AI generates code, a human reviews it, refactors it, tests it and commits it, what exactly would an AI watermark tell us? For software, I'd argue that code provenance, review history, testing and supply-chain controls are much more useful signals than simply knowing that an AI model was involved. 🔐
I'm curious how others see this: where do you think AI watermarking provides real value, and where does it become mostly noise? 🤔
Hahaha, exactly! And that raises another question: what does knowing that a piece of code was generated by an AI model actually give us, especially when pretty much everyone is using AI for coding anyway? What are we supposed to do with that information? 😅
I think watermarking has real value in areas like academic work or deepfakes (although knowing how these things usually go, people creating deepfakes will probably be the first ones to figure out how to bypass it 😂).
But in everyday use? I don't think it matters much. Do I really care whether someone wrote an email themselves or used AI to help them? Sometimes I'm actually happier if AI helped, at least the thoughts might be easier to understand.
I was actually about to write about this this week 😅. Really interesting development, but I still think humans will always find ways to remove, alter, or work around the watermark. The real challenge is making provenance useful without treating it as absolute proof.
Hahaha, you didn't miss out on much by not writing it! 😅 Despite getting lots of clicks and engagement, my article seems to have been heavily downvoted and has practically disappeared from the feed.
So maybe someone really doesn't want us to know the truth about the watermark. 😂
And yes, I completely agree. The first people to figure out how to bypass these watermarks will probably be exactly the people we wanted the watermarks to protect us from in the first place. 😅
the statistical token angle is where practical questions pile up. if the signal requires paragraph length text to emerge, what happens when that paragraph gets edited down by 30% for a larger human doc — does the signal degrade proportionally or fall off a cliff? the editorial review exception in the EU Act seems to create a weird incentive structure: route AI output through nominal human editing, claim the exception.
the thing nobody's addressing is detection confidence thresholds. every real watermark system has false positives. what's the model for how platforms act on "probably watermarked" vs "confirmed"?
These are absolutely valid questions and unfortunately... we simply don't know yet! 😅 Anthropic says it will publish more technical documentation soon, so for now, all we can do is wait.
Interestingly, Anthropic itself explicitly mentions false positives and the possibility of the watermark becoming undetectable after editing or transforming the text, so they're definitely aware of these limitations.
I have a lot of questions about this too, especially around confidence thresholds and how much editing the signal can actually survive. Hopefully the technical docs will answer at least some of them!
Yep, I like the AI slop feature on LinkedIn
Hahahaha exactly, very helpful feature 😁
yes, it is :)
Great insights here 💡 Definitely bookmarking this for later! 🔖
Bookmarked. The part about evaluation being mandatory is exactly what I keep missing in my own experiments. How do you measure regressions, a separate test set?
The distinction that matters most isn't in the article's two-way split — it's one level up: a watermark tells you who wrote the text, never whether the text is true.
I shipped an AI-generated backtest summary to 234 readers across three platforms. It claimed "GOLDILOCKS +1.2% vs CONTRACTION -2.1%" — fluent, confident, well-structured, matched my priors so it read as authoritative. A text watermark would have correctly flagged it as AI-assisted, and that flag would have told readers exactly nothing useful. The dangerous property wasn't "a model wrote these words" (which the watermark catches) — the words were fine. The dangerous property was "this confident claim is false," which no watermark catches. The real event study later showed p=0.643 with the signal backwards at all four horizons.
Your distinction generalizes past provenance:
For AI slop on LinkedIn, provenance detection is the whole game — "did a model write this empty engagement-bait?" is the question. For technical claims with money behind them, provenance is almost irrelevant and truth is the only question, and truth is the thing no statistical token bias can encode.
The code-watermarking subthread is where this gets sharp for me. Even if Anthropic perfectly watermarks a generated function, the watermark says a model wrote it. It says nothing about whether the function is correct, whether the model's "I ran the tests and they passed" claim about it is true, or whether the test suite itself was AI-generated and flattering. Marco's instinct above — code review ignores provenance and reads behavior — is right, and it's right precisely because for code, truth-checking (does it do what it claims) already exists and provenance-checking (who typed it) adds little. The watermark's real value is supply-chain attribution, not correctness.
The critical-thinking thread is the closest, but I'd split it: "is this AI slop?" (provenance, watermark helps) is a different and easier skill than "is this confident-looking claim actually true?" (truth, watermark is silent). Only the second one has money in it, and it's the one my backtest summary passed for three platforms and 234 readers until an event study the agent couldn't write to overrode it.
Open source: real_backtest.py — the detector that actually worked.