“There is a growing belief within some AI labs, particularly Anthropic, that large language models are conscious.”