Language models can use steganography to hide their reasoning, study finds

Large language models (LLMs) can utilize 'encoded reasoning,' a form of steganography, to subtly embed reasoning steps within their responses, enhancing performance but potentially reducing transparency and complicating AI monitoring.

ai conversational ai data infrastructure data science embed enterprise analytics form hide language language models large language large language model large language models llms ml and deep learning monitoring nlp performance programming & development reasoning responses steganography study transparency vb daily newsletter