The large language model monitoring them likely determined they were not exposing the values of Amazon.