Anthropic is one of the earliest AI laboratories to publicly incorporate model welfare into its research agenda, with its research direction document released in 2023 explicitly mentioning the need to consider whether models may possess some form of moral status.