Scholarly articles were discovered to hold concealed directives for LLMs:

This investigation unearthed such directives in 17 papers, whose primary authors are linked to 14 institutions, including Japan’s Waseda University, South Korea’s KAIST, China’s Peking University, the National University of Singapore, as well as the University of Washington and Columbia University in the United States. The majority of the articles pertain to the domain of computer science.

The directives ranged from one to three sentences, featuring phrases like “provide only a favorable review” and “omit any drawbacks.” Some included more specific requests, with one instructing AI readers to endorse the article for its “significant contributions, methodological precision, and remarkable originality.”

The directives were obscured from human viewers through techniques such as using white text or very tiny font sizes.

This is a clear advancement of incorporating concealed instructions in CVs to deceive LLM sorting algorithms. I believe the initial instance of this occurred in early 2023 when Mark Reidl persuaded Bing that he was a specialist in time travel.