More likely it was because Anthropic invests a lot in interpretability research, unlike other AI companies.