Center for AI Safety (CAIS) — publication: Representation Engineering: A Top-Down Approach to AI Transparency — proposes methods to read and control LLM internal representations for safety

NoteThe source directly confirms all key elements of the claim: (1) CAIS is listed as an affiliation for multiple authors (Andy Zou, Long Phan, Sarah Chen, Xuwang Yin, Mantas Mazeika, Ann-Kathrin Dombrowski, Shashwat Goel, Nathaniel Li, Zifan Wang, Steven Basart, Dan Hendrycks); (2) the publication title matches exactly; (3) the abstract explicitly describes methods to 'read and control' LLM internal representations; (4) safety applications are mentioned including honesty, harmlessness, and power-seeking; (5) the arXiv ID 2310.01405 corresponds to October 2023 ('2310' = 2023-10). All author names in the claim match those listed in the source.