The AI features of Boldcaster run on Mistral AI, a provider based in France that processes on servers within the European Union. No US provider is involved in analysing strategy, numbers, goals or personnel data.
The use of client data to train AI models is disabled in the account settings at Mistral and is contractually excluded. The Processor does not train any models on client data itself and does not provide client data to third parties for that purpose.
Input is retained by the AI provider for a maximum of thirty days for abuse detection and then deleted. The Processor has no access to other clients' conversations at this provider.
When drawing up reports about team members, names are replaced by a neutral label before sending. The AI provider therefore receives no identifiable assessments of individual team members. The translation back to the real name only happens within the application.
AI connection at the Controller's request. The Controller can enable, per organisation, a connection that lets authorized users operate Boldcaster from an AI application of their choice, such as Claude, ChatGPT or Vibe. This connection is off by default. When it is on, the Processor provides data to that application only at the request of an authorized user, within that user's rights and within the parts the Controller has enabled.
The provider of such an AI application is not a sub-processor of the Processor. The Controller or the user chooses and contracts that provider, and processing by that provider falls outside this agreement, including where it takes place outside the European Economic Area. The Processor records every request made through the connection in a log the Controller can view. A user and the Controller can revoke access at any time.