Confidence Levels: Ensuring Privacy and Trust

The Confidence Level feature in AI Studio is designed to help you use only those Large Language Models (LLMs) that meet your privacy requirements for your specific use case.

For instance, if your work involves processing highly sensitive data, you can configure the confidence scheme to trust only self-hosted models. This ensures that sensitive information remains restricted to local environments.

Note: The confidence levels used here are for demonstration purposes only and should not be used as guidelines for individual confidence levels.

Customizable Confidence Schemes

Pre-installed Levels: AI Studio includes predefined confidence levels based on the country of origin, helping you evaluate the trustworthiness of models at a glance (e.g., self-hosted, US-based, or EU-based models).

Custom Configurations: You can create individual confidence schemes tailored to your preferences, categorizing models based on their provider (e.g., self-hosted, OpenAI, Anthropic, etc.).

Default Settings

By default, AI Studio does not enforce a minimum confidence level, shows the different confidence levels, and and trusts providers from both the US and Europe app-wide.

Follow these steps to adjust the default confidence settings in AI Studio:

1. Go to Settings

2. Select Configure Providers

3. Enforcing a Minimum Confidence Level

You can enable the option to enforce a minimum confidence level across the app if needed. When this feature is enabled, only models meeting or exceeding the specified minimum confidence level will be available for use in chats or assistants, ensuring that all interactions align with your privacy and trust requirements.

4. Select a Confidence Scheme

      4.1. Use Pre-installed Levels

            a. Open the Select a confidence scheme drop-down menu.

            b. Choose one of the pre-installed schemes from the list.

            

      4.2. Create a Custom Configuration

            a. Open the Select a confidence scheme drop-down menu.

            b. Click Configure your own confidence scheme.

            c. Customize the Confidence Level for each LLM Provider by selecting an option from their respective drop-down menus.

            Important: Once you set the Confidence Level for a provider, you must select one of the levels. It is not possible to revert to Not Yet Configured.

            

5. Viewing the Confidence Card Once different confidence levels have been configured, you can review a Confidence Card for each model directly within chats or assistants. The Confidence Card provides an overview of the model’s assigned confidence level, helping you quickly verify its compliance with your trust and privacy requirements.