The EU’s AI rules are having an impact.
Archived version: https://archive.is/20260811122405/https://www.theverge.com/ai-artificial-intelligence/977823/anthropic-claude-ai-watermarks-c2pa-text-images
The EU’s AI rules are having an impact.
Archived version: https://archive.is/20260811122405/https://www.theverge.com/ai-artificial-intelligence/977823/anthropic-claude-ai-watermarks-c2pa-text-images
17 Comments
mrmaplebar@fedia.io · 13 pts · 27d
Not enough. AI generated content should be required by law to contain both visible and invisible watermarks.
floofloof@lemmy.ca · 11 pts · 27d
You can't enforce that with text though, because people will just delete the words. They use a subtle statistical signal because people can't just spot it and delete it. Unfortunately it raises a problem: either they have to make the signal a secret, in which case only Anthropic and those contracted to keep the secret could detect it, or it's not, in which case people can use a detector and some other AI to discover how to eliminate the signal. I expect Anthropic would try to keep control over the detection secrets and provide a detection API for third-party software to use.
Ek-Hou-Van-Braai@piefed.social · 5 pts · 27d
Unless you're using AI to write books, I can't imagine how you'd implement some secret watermark
floofloof@lemmy.ca · 3 pts · 27d
As I understand it, you're right that you need a good length of output to be able to detect the watermark, because only then can you see the statistical effect with confidence. And in code there are usually various options for how you get something done, and they could watermark generated code by adding a distinctive pattern to its preferences for certain constructs over others. But again, it would have to be subtle, so you'd need a large enough sample of its output before you could see the effect.
Ek-Hou-Van-Braai@piefed.social · 2 pts · 27d
It should be fairly easy to stop this detention even in code.
You just need to set very strict "rules" the AI must follow when writing code.
Variable and method names etc. Must follow a specific formula.
Have a different AI write your docs etc.
If you're just vibe coding everything, the watermark would work. So maybe a good thing. But it'll be whack a mole, and if you're careful I'm sure you can stop the detection
schnurrito@discuss.tchncs.de · 1 pts · 27d
You know, the evil bit used to be a joke, now people unironically suggest equivalent things...
TrickDacy@lemmy.world · 0 pts · 27d
I'm so silly. I saw this and thought, oh cool a teeny tiny surprising win. Guess I should only be happy with the destruction of all LLMs.
SnailMagnitude@mander.xyz · 10 pts · 27d
This seems a riot, trying to sneak in invisible copyright really does seem like the copyright system is well and truly fucked.
It arrived to try and help with the deluge of slop from Guttenberg nonsense but the whole idea, and copyleft, seems to be imploding at the moment.
godsammitdam@lemmy.zip · 7 pts · 27d
So then they can lie and say real pictures have the watermark?
So they can lie and say fake images don't have the watermark?
So the AI looking for the watermark can hallucinate?
Yeah, I'm sure that's super reliable.
undefinedTruth@lemmy.zip · 6 pts · 27d
Can someone explain to me how is it possible to watermark text? Because it makes no sense to me.
schnurrito@discuss.tchncs.de · 3 pts · 27d
Nor to me - but it doesn't make a lot of sense with images either.
Text is literally just an array of numbers (Unicode character points); where do you put "invisible" watermarks there? Images are, likewise, a three-dimensional array of pixel color values; while it's possible to very slightly vary those color values and call that "watermarking" (similar concept: steganography), how does that not get lost even unintentionally on the first lossy compression step?!
MattR@feddit.org · 1 pts · 26d
There are multiple major flaws with watermarks for texts:
Promethiel@lemmy.world · 1 pts · 27d
Google already does this with Gemini.
Not only does the technical side work very intermittently at detection when you do carry over the same file, but with an "invisible" watermark, it can be bypassed entirely as easily as with a screenshot.
Double_A@discuss.tchncs.de · -4 pts · 27d
And who will care about those? The slop consumers will still consume...
BakedCatboy@lemmy.ml · 7 pts · 27d
Personally I would love a browser extension that detects these and adds a badge to AI content.
Solumbran@lemmy.world · 5 pts · 27d
People who want to know if it is slop or not quickly?
historicaldocuments@lemmy.world · 1 pts · 27d
If it's there and has been there, then there's always a chance it can be used later: