Toggle light / dark theme

AIs Can Store Secret Messages in Their Text That Are Imperceptible to Humans

A recent study found that LLMs are incredibly good steganographers, able to obscure their own thought processes through “encoded reasoning.”