Get the latest tech news

AI watermarking could make LLM guardrail adherence unpredictable — and that could be a big problem for the EU AI Act


Researchers have revealed how AI watermarking could actually change how an LLM behaves, making it less secure.

None

Get the Android app

Or read this on TechRadar

Read more on:

Photo of LLM

LLM

Photo of big problem

big problem

Photo of EU AI Act

EU AI Act

Related news:

News photo

LLMs respond differently to harmful prompts when AI watermarking is used

News photo

How to Write with an LLM

News photo

PrismML hopes its tiny LLM will change how we all use AI