Enterprise admins need transparency and control over the AI tools running in their environment. With that in mind, we're introducing model selection for AI Classification policies, giving admins control over which AI model powers their classification workloads. You can now pick between two paths directly from the Admin Console:
- Default (GPT): The model that AI Classification is built and optimized for. If you're looking for the best out-of-the-box classification performance, this is your path.
- Compliant (Gemini): Designed for organizations with stricter regulatory requirements. Gemini offers superior compliance rigor, making it the right choice for FedRAMP environments and regulated customers who need explicit, auditable control over which AI models process their content.
This decision will be applied consistently across live classification, fallback agents, Test and Iterate, and any future applications of AI Classification.
To learn more about Box Shield Pro and AI Classification, look here.