FYI this is about humans using AI tools for better steganography (the practice of hiding secret messages in non-secret data). The article title is rather clickbaity, as (at least to me) it conjures up images of rogue AI communicating without humans knowing about it.
Also, the article describes the method as “perfectly secure.” The researchers do use this phrase, but it has a specific technical meaning and does not necessarily imply that this method is now and will forever be undetectable. From the linked research abstract:
In this work, we show that a steganography procedure is perfectly secure under Cachin (1998)'s information theoretic-model of steganography if and only if it is induced by a coupling. Furthermore, we show that, among perfectly secure procedures, a procedure is maximally efficient if and only if it is induced by a minimum entropy coupling. These insights yield what are, to the best of our knowledge, the first steganography algorithms to achieve perfect security guarantees with non-trivial efficiency; additionally, these algorithms are highly scalable.
The article does not mention whether these claims have been verified by other researchers. I am not knowledgeable enough to evaluate their research and their claims, so I am not trying to dismiss their finding. It seems like a new, interesting approach to steganography. But I think it’s important to keep a little healthy skepticism when reading articles like this, especially with all the AI hype going around.