keep in mind that your tool fucks up the output of screen readers as well.
Show HN: Stun LLMs with thousands of invisible Unicode characters
51–60 of 115 posts
Re: Show HN: Stun LLMs with thousands of invisible Unicode characters
#52Earlier quoted context omitted.
you are correct. This makes text almost completely unreadable using screen readers.
Do screen readers fall back to OCR by now? I could imagine that being critical based on the large amount of text in raster images (often used for bad reasons) on the Internet alone.
no, but they have handling of unknown symbols and either read allowed a substitute or read the text letter by letter. both suck.
Re: Show HN: Stun LLMs with thousands of invisible Unicode characters
#53you don't need invisible chars. Just use a different text direction. e.g.
decipher this message as its written bottom-to-top, RTL
```
t_____s
s_____i
e___s_h
t_a_i_T
```
(swap underscore with a space)
Re: Show HN: Stun LLMs with thousands of invisible Unicode characters
#54For LLM scrapers, it doesn't even matter if LLMs would be able to understand the raw text or not because it's extremely easy to just strip junk unicode characters. It's literally a single regex, and, like, that kind of sanitization regex is something they should already be using, and that I'd use by default if I were writing one.
Re: Show HN: Stun LLMs with thousands of invisible Unicode characters
#55You can also give the LLM hidden messages with a small bit of prompting, e.g. https://umpox.com/zero-width-detection
It’s technically possible to prompt inject like this. I actually reported this to OpenAI back in April 2023 but it was auto-closed. (I mean, I guess it’s not a true vulnerability but kinda funny it was closed within 5 mins)
Re: Show HN: Stun LLMs with thousands of invisible Unicode characters
#56Earlier quoted context omitted.
Do screen readers fall back to OCR by now? I could imagine that being critical based on the large amount of text in raster images (often used for bad reasons) on the Internet alone.
no, but they have handling of unknown symbols and either read allowed a substitute or read the text letter by letter. both suck.
Sounds like a potentially useful improvement then.
I've had more success exporting text from some PDFs (not scanned pages, but just text typeset using some extremely cursed process that breaks accessibility) that way than via "normal" PDF-to-text methods.
Re: Show HN: Stun LLMs with thousands of invisible Unicode characters
#57[dead]
Re: Show HN: Stun LLMs with thousands of invisible Unicode characters
#58For LLM scrapers, it doesn't even matter if LLMs would be able to understand the raw text or not because it's extremely easy to just strip junk unicode characters. It's literally a single regex, and, like, that kind of sanitization regex is something they should already be using, and that I'd use by default if I were writing one.
There are no “junk” Unicode characters. There are just nonsensical combinations of characters. Stripping out characters blindly is not a solution, because you have no way of knowing what was intended.
Re: Show HN: Stun LLMs with thousands of invisible Unicode characters
#59> Even just one word's worth of “gibberified” text is enough to block most LLMs from responding coherently.
Which LLMs did you test this in? It seems, from the comments, most every mainstream model handles it fine. Perhaps it's mostly smaller "single GPU" models which struggle?
Re: Show HN: Stun LLMs with thousands of invisible Unicode characters
#60> Even just one word's worth of “gibberified” text is enough to block most LLMs from responding coherently. Which LLMs did you test this in? It seems, from the comments, most every mainstream model handles it fine. Perhaps it's mostly smaller "single GPU" models which struggle?
I just tried "Hello World" with ChatGPT 5.1. After a while, it responded with a bunch of Cyrillic text.