Research documenting how adversarially crafted tool outputs can establish false premises in language models, leading to …