Anthropic is different from OpenAI in that instead of people in Kenya manually rating the outputs for toxicity or whatever, they have this constitutional document based on the Universal Declaration of Human Rights and all that, and the Sparrow Principles, actually, from DeepMind. But it seems that it still has bias. And for example, most of the users, actually, are not American, but especially for people who are Muslim, who use either OpenAI or a cloud from Anthropic, there is a lot of implicit bias just based on the ways it’s used and so and so. So, we work with them.