Hacker News

Also why color laser printers refuse to print b/w documents when they run out of color toner since they can no longer print the yellow DocuColor tracking dots.

Oh that makes it totally ok then...

And? Where are we going with this whataboutism?

Probably just means they haven't (yet) figured out how to do the watermarking carrying enough bits of data to encode a GUID.

At least, according to the very trustworthy Claude.

Exactly it is a legal requirement from lawmakers who either don't care or understand privacy implications to implement provenance metadata. California's act came into effect August 1 https://sd13.senate.ca.gov/news/press-release/august-3-2026/...

Well I will never ever buy something from Microsoft again because they do hate their customers.

I own an Xbox with an family account. Microsoft's account management is terrible. Their customer support is terrible (computer says no). Never again.

How many Fortune 500 companies require you to buy your own Windows laptop as opposed to them providing one?

Also, if your laptop is work provided, your laptop is going to have little to no privacy no matter what.

If you work at a company and it's a company device, then its not your privacy to worry about - you don't use your work device for personal matters, and the concerns are someone elses.

IBM, SAP, Capital One, GE, Cisco, Intel, Deloitte offer a Mac/Windows choice program. Better than just one choice, and arguably a bit better of a privacy record than MS (especially of late)

"privileged take"?, sorry I am not from the US, so I don't compete in the sorry-o-lympics.

Nothing is stopping you from using it for free at home, everyone is privileged in that regard.

And it was obvious to everyone else, I was talking about decisions you can take yourself. I'll try better in the future to be hyper explicit about this, so people who are looking for a reason to feel hurt will have a harder time.

They have only themselves to blame for this. They've spent decades developing technology to spy on and micro-regulate what their captive customers can do with their property. Now that they have this technology, it's only logical that they be held responsible for misuse.

The only (and inevitable) solution at this point is for big tech to get micro-regulated like the banking industry. The same mindless greed that drove them to take control away from their customers will be responsible for the same control to be taken away from them.

The guardrails that prevent generating adult/explicit/whatever content are a completely separate step. This has nothing to do with that.

This has nothing to do with generating harmful content and everything to do with fascist surveillance. This is an OS level functionality that get utilized before AI even existed.

> Not that I love Microsoft, but they would get a lot of heat if they let anybody generate harmful content as well.

Nonsense. They let anyone draw any pic they like.

You know, you're only supposed to drink the kool aid, not chug it by the gallon.

The article says that once converted to BMP all the metadata gets removed....so on linux:

convert file.jpg file.bmp; convert file.bmp file.jpg

There are several paragraphs where output looks very AI-like (and Claude flavored one at that), e.g.

> In other words, “generated locally” does not mean that the complete operation is local. Microsoft receives and moderates the prompt, then issues the unique GUID that Paint embeds into the locally generated image. Paint also sends the previous promptGenerationId as lastPromptGenerationId with its next moderation request, allowing successive requests to be linked explicitly.

> That relationship is important. C2PA calls this a soft binding: a value derived from, or embedded into, the content so that the content can still be matched with its provenance record after the file-level manifest has been removed. For a watermark soft binding, the value is the watermark’s content identifier. Microsoft cryptographically signed this assertion.

> After an AI result is applied to the Paint canvas, the available formats are still restricted to PNG, JPEG, GIF, and Paint’s own .paint format. BMP—the classic Paint format—is conspicuously absent.

Personally it didn't bother me too much.

They're saying that the blogpost is at least partially AI-generated.

Some people are suggesting systemd is spying on you because of /etc/machine-id

> It's very likely your printer is secretly adding marks to the page

It's most likely how the FBI caught NSA leaker Reality Winner:

> Both journalists and security experts have suggested that The Intercept's handling of the documents, which included publishing the documents unredacted and including the printer tracking dots, was used to identify Winner as the leaker.

https://en.wikipedia.org/wiki/Reality_Winner

>Then they link them to your MS/Apple account, IP addresses, and the extended social-graph of all computers that were ever in the same room or shared the same bluetooth speaker.

Did it get revealed that Apple did that? This is news to me

So what you're saying is we should use linux?

and then in few years a new mono-mustachioed or mono-browed leader will emerge who declares memes a blasphemy punishable by death and will call up all the telemetry gathered to punish all involved.

No, because the GU in GUID stands for globally unique. Adding it to a second image would cause a detectable collision.

Also, C2PA, another technology mentioned in the article, means tampering is easily detected.

Not even. They log the IP and your ISP (or VPN provider!) knows the name/address of the person using that ip at that time.

I was forced to create multiple as part of the mojang migration - truly a nightmarish experience to use their AuthN, haven’t seen something quite as bad since

Assuming you are on windows, you probably have

It does kind of say: The GUID is coming from the moderation endpoint, which is hit when you generate a local or cloud AI image based on your prompt. If there is no prompt, there is no endpoint, and likely no GUID.

Obviously Paint could have been watermarking prior to AI though, but this specific AI watermarking appears to be only that.

While it does make a difference, their willingness to quietly integrate watermarking features into what people saw as simple apps for doing simple tasks is unnerving. It only takes a small change for them to start baking your identifiable information into all images they edit, or some government politely asking them to do that.

I wonder if in ten years we'll have a horrifying world where everything that leaves a machine is imprinted with its permanent identifier. Every file comes with a verifiable history of who created it, what computers it passed through, who made edits. We're closer to that world than we think.

Windows 11 effectively forces users to register with a Microsoft account. Once that's established, all it takes is for an unaware user to fill out an e-commerce form and save an address for auto-fill.

> Make sure you register mailing addresses with your credit card institutions

Sorry for a dumb question but what would one use as a mailing address? Rent a PO box, or use a mail forwarding service, something like that?

This is broadly impractical for the average citizen. A scalable solution would be to make this sort of thing illegal.

> IANAL but they are not a government or financial institution and do not have a right to that information.

As soon as you purchase something from Microsoft - e.g. your Office 365 subscription - they have at the very least your billing address on file for the credit card.

add your IP, location, provider, computer specs, dimensions, screen info, nearby devices, etc etc etc

Ain't nobody anon anymore thanks to the image recording GPS radio in the pocket.

Watermarking is just one puzzle piece in the surveillance state; the legislation around it another.

You can incentivize cooperation without having to compel action with a warrant. This sort of corrupt quid pro quo is quite common, in other countries at least...

Why would the EU possibly object to this? It is exactly what they want.

> This really needs to be hit with the GDPR hammer. Microsoft have not obtained consent for this.

At best they'd just disable it for EU... assuming they didn't successfully argue "it was in the ToS ..."

Watermarking ai generated stuff is mandated by EU

a GUID isn't an indicator, it's a fingerprint.

so unless you want to draw a distinction between 'user' and 'machine' , yeah it is for identifying users.

to believe otherwise, especially with Microsoft involved, would be incredibly naive to their history.

Article literally says it adds a guid, not a binary field indicating that the image is ai generated.

> I think there is a good argument for being able to distinguish AI generated image and video...

Neat but that's not what's being built here. What's being built is "we can trace back this content to who made it" which is bad. Doesn't matter if today that it's limited to AI generated content. Won't be tomorrow. Your devices should not act against your best interests. No cop in my pocket please.

Paint.net is named after the dot net framework and is not referring to a URL or the internet as I understand it.

I was stating that MS Paint mis categorized a Print Screen screenshot as AI generated when pasted. Which pushed me to install PaintdotNet onto my laptop instead.

They aren’t saying that, paint.net is an open source application, Paint is the descendant of MSPaint.

Yeah, I've been working around that for now by uninstalling notepad. Which actually uninstalls the store app, which in turn exposes the actual notepad executable that ships with Windows. Really confusing, and will probably break in some future version of Windows.

I don't understand why they didn't just stuff all the AI things and new features in Write, where it belongs…

It automatically saves and reopens whatever you previously pasted in there. Quite a shock the first few times it happens. It's not a safe temporary holding area anymore.

I don't use Windows, but I have to say that TextEdit on the Mac is absolute crap. Literally every default is wrong. Text is too small, looks like crap, cut-and-paste always brings absolutely crappy styling with it, its autocomplete and autocorrect is bad and wrong and annoying. No one at Apple uses this application with any regularity and it shows.

I wanted dark mode, tabs, and autosave, and now I don't have to use a different app to get those, so that works out great for me. Not sure what "breaking changes" even means for that though. Did it suddenly stop being able to open text files? No, it still just works.

For a long time I used metapad as a notepad replacement https://en.wikipedia.org/wiki/Metapad (I even renamed the exe to notepad). Very fast and minimal.

Uninstall bloated notepad will revert it to the default version.

Doing so with MSPaint removes the app, so you need to find a clean older version..

I've been using ComfyUI to do fine background and watermark removal. Might take a look into that as an alternative. Run Flux.2 Klein or Comfyui's own workflow.

Pixelation was a feature back in the 90’s and 2000’s.

But yea, I guess it’s a bug in 2026.

If you don’t mind CLI, rembg is a good option for background removal. It lets you swap between different models depending on the type of image so you can use something like `birefnet` for photorealistic images and `toonout` for illustrations.

https://github.com/danielgatis/rembg

I think their point is that paint shouldn’t have any GenAI features (or any new features other than compatibility)

It's a problem on so many levels.

We've probably lost the fight over GenAI marks, at least for now, and I do understand the need for them. However the same technology that can embed an invisible watermark that survives a round of photographing, printing, crumbling and scanning back - can be used to embed more information than just "it's AI generated". Encoding GUID is just a harbinger; as a next step, why not encode the app that produced the image? Also a hash of its license key? Hell, what's stopping an app from doing it itself, watermarking its own window before passing the pixels to OS for compositing?

"Robust GenAI watermark" today is "robust general-purpose DRM watermark" tomorrow :).

I don't think adding an identifier which can most likely be mapped back to user is "standard mark".

It doesn't explicitly say it, but there is hardly any other reason why it would add server generated ID (from authenticated session) to the saved file.

Cropping and editing screenshots (eg add red circle) in a very lightweight application.

Other image editing applications can do all of that, but they take 10 years to launch (during which they will ask for focus) and 2 GB of RAM.

Used to be you saved and shared screenshots by pasting into Paint and saving it. Printscreen didn't save a file and only some programs (Skype) let you paste the screenshot directly from the clipboard. That's how I remember it anyway.

The utility of Paint is that lightweight software is an absolute fucking godsend (which, Microsoft being Microsoft, is on a mission to ruin). I have all manner of professional tools - Photoshop, Aseprite, Manga Studio, etc, and if the edit is something that can be done in Paint, I will use Paint over any of the others, because I'll be done editing before I've even managed to open the file in another program.

Yeah you can do that, and it's a crime

Virtually all color laser printers and copiers.

Now I need to see if agentic reverse engineering of printer firmware can actually remove that "feature" for good.

It was about money anti-counterfeiting.

https://en.wikipedia.org/wiki/Printer_tracking_dots

Or EU money.

https://en.wikipedia.org/wiki/EURion_constellation

There is no robust-without-qualifiers watermarking. There is only robustness against common operations but as soon as someone is actually trying to remove the watermark they can at the very least use exactly the same technique to embed random noise instead to overwrite your watermark.

tbh so is circumventing those dots.

It doesn't make it any better

Anthropic's watermarking technique is possible, and worryingly could effect low frequency choices (eg as I said, with object placement or style). However, this has to be a generation time decision, you can't really do that with a small edit. In fact small edits tend to be detectable spectrally.

Don't care that you don't care. This distinction does matter to a lot of people, because to many people there's a huge difference.

I'd never play one of those games but the excuse for the kernel modules spying on you is usually anti-cheat not DRM as they are online games.

There are droves of people petitioning Valve to add kernel anti–cheat to CS2.

I have the same issue on Windows 11. It's been bugging me to press a button or "check my drive" on every startup for at least half a year, no matter how often I let it run the check...

Why have you linked a bug report for a niche userland application that may be causing disk corruption as evidence against linux reliability? That seems like a pretty uninformed conclusion.

Do those users use the OS that randomly takes hours to shut down because it desperately needs to install updates right now - and then it won't shut down at all and reboot instead. And even if it doesn't do that shutdown takes ages and you need to attend it and can't just walk away because the OS will drop you in a logged in desktop when some random application decides to interrupt the shutdown process.

I don't know why you'd choose one of the smaller OSes, I'm required to use Ubuntu at work and it just works fine for me, every day.

It's also a bit rich especially to complain when Microsoft also suffers from forced autoupdates and the like.

The best response is always:

"That's not up to you."

When people claim they have nothing to hide, always point out that's not up to their determination. That freaks them out, it disarms their shrink-from-confrontation move.

And you can point it out super fast, in a plain six word statement. No need to launch into a deep discussion unless prompted.

Alternatively, simply respond with: yes you do. When they reply: "what?" - "every single thing you have ever done wrong across your entire life." Everybody has done something they would prefer to keep hidden, the cowards just lie about it.

Essentially any Linux distro protects privacy. You'd kinda have to go out of your way to find one that doesn't, because there's no online account connected to your install for any of them.

If you're reasonably technical, you can make nearly any use-case work on nearly any distro, but if you have choice paralysis, my top recommendations would be CachyOS if you plan to play games, and Mint otherwise.

Personally I'm happy with EndeavourOS. I picked it to find a general purpose distro similar to the Steam Deck (KDE and arch based) but with a more user-friendly installer.

Pretty much all of them. None require online accounts or anything like that. Browser choice is probably more important.

For something that just works I recommend Fedora or Ubuntu. It's what I'd put on my mom or wife's PC.

Currently using Omarchy on my laptop though which is a tad more exotic, it's Arch + Hyprland and is pretty polished for that stack (at least the current version, 4.0) but still a tiny bit of jank.

omarchy.org

Randomizing bit 0 might not be enough but if you know the algorithm used (and you do in this case because it is applied locally) then you can always replace the embedded data with random noise.

No, I wasn't suggesting that. I was saying that if there was digital art (human made) aesthetically significant that a curator would want to display it in a museum, Microsoft's GUID supplied to a data collection agency would make it possible to retrieve if ever/whenever that data were declassified (assuming it isn't purged)

It's possible a very bad curator with a terrible taste in art might select slop to display, but I was refering to "fine art" or at least finer art that is digital.

So as usual, exactly the things you want it to work on it won’t.

It actually might make the new horrible world even worse. Imagine the populace getting used to a AI image detector flagging things as fake using this fairly easily defeated GUID marker system. Most people are just making memes or cat videos and don't even try to remove this so eventually the populace starts to believe these things actually work.

Now some one slightly more sophisticated starts creating deepfakes of a woman and uploading them or fabricating video of an political event without this marker. The subject protests it's fake and AI generated but a loud majority of ignorants feed it into Microsoft AI detector and call you a liar and say it's confirmed real. Most people don't know any better and eat it up because a computer said so.

I guess that makes sense. So import an image of the mandelbrot set you generated and save it, and see if any pixels change.

It's already illegal to not watermark an AI image.

Without access to the code, I imagine it'd be extremely difficult even with unlimited API calls to the watermarker.

I've done this btw. I went for the Pixel Camera app since they were the ones bragging the hardest about their "security". Writeup + PoC should be dropping some time tomorrow. Despite 90+ days from initial report, it remains unpatched.

Some proof: https://verify.contentauthenticity.org/?source=https://retr0...

I could also paste a privkey + cert chain in here but el goog's lawyers might not like that.

Imagine today where a photo is submitted as evidence and the court believes it even without signatures.

I was thinking any photo created with a camera should be signed. Why we don't have that in 2026 is beyond me.

But what you're talking about is the generative aspect of these photos likely expanding over time. We're seeing that today with the ultra zoom features on some cameras regenerating objects (and especially text). Without the user doing anything the phone will generatively fill in detail, most worryingly text and people. Then there's the Samsung moon issue - taking a photo of a pixelated printout of the moon caused Samsung phones to generate a new image of the moon.

In the imaginary dream world that Adobe, Google et al live in, the final file does indeed have a signature.

Each piece of software in the chain must use TPM-like technologies (yes, even GIMP) to make sure it's running a "legitimate" build of the software, on "legitimate" hardware, and re-sign the file at each step along the way (using keys provisioned during some flavour of remote attestation flow, or using a RA-authenticated remote-signing oracle).

The final file embeds every preceding manifest, so you can "verify" all the way back to the original.

If this all sounds patently unworkable, that's because it is.

Presumably the OP is proposing something like a TPM attached to the image sensor that signs the sensor output or something like that. You can’t sign it because you can’t get the key out. The key could be per-camera and be a published list.

I suppose a dedicated fraudster could still stage an appropriate scene. An appropriately lit matte image might even suffice.

Keys could be stored in something like TPM on Camera, and could sign the image. The key could then be verified from the camera itself to prove the authenticity of the image.

If we as a society deemed it necessary, the camera manufacturer could also provide a list of keys for devices they have manufactured. And an image/key could be provided, and the manufacturer could verify the authenticity that way.

The TPM signing could be tied into the sensor hardware itself, making it difficult, but not impossible, to sign arbitrary images with the TPM.

Or, you know, using the totally-real-picture camera to take a photo of an AI-generated scene?

Asymetric keys

Well that was fast!

> build their own photo-sensor simulator and plug _that_ in to the camera's motherboard to feed it a "real" image?

Already been done!

https://hackerfactor.com/blog/index.php?/archives/1102-C2PA-...

TL;DR: Strong cryptography over untrusted data does not make the data more trustworthy.

Stalinist Russia? Stalin would be absolutely dumbfounded by the level of tracking common in the west today.