Techradar says it’s based on the Mistral 7B large language model. But they should definitely disclose that kind of information. It’s important to know how a tool works and what kind of mistakes, biases etc are to be expected when using it for important communication.

They is no chance they are the one training it. It costs hundreds of millions to get a descent model. Seems like they will be using mistral, who have scrapped pretty much 100% of the web to use as training data.