Hacker News Reader

My security advice is: if you are feeding documents to your AI whose authors might have adversarial intent, transform them into photos and run OCR on them. Yes, it is wasteful and slow, but it is a guaranteed way to get rid of invisible (meta)data.

Granted, the OCR does not have a perfect 1:1 correlation to what the human eye sees, but IMHO as of November 2026 it is a fairly good approximation.

View on HN

Also LLM based vision models are shockingly better than OCR in almost every OCR related task. Very little tradeoff on quality, only speed.

Right, I didn't realize OCR is not used anymore when it comes to LLMs. Yeah, vision models, that's what I meant.

Well, OCR is the name of an application, just like LLM. If you take the same neural net that runs on LLM and use it to parse the photo of a document, it's an OCR.

I think you are confused about what "LLM" means [1]

1: https://en.wikipedia.org/wiki/Large_language_model

It’s a way to get rid of some attacks but shouldn’t be seen as a panacea.

Reply on news.ycombinator.com