"Researchers at the A.I. company Anthropic claim to have found clues about the inner workings of large language models, possibly helping to prevent their misuse and to curb their potential threats." # Description used for search engine.
Read more here: External Link