OpenAI high risk use case mitigation requirements
In this page, learn about the mitigation requirements for OpenAI high risk use cases. You are responsible for implementing these requirements when using OpenAI models on Databricks
Mitigation requirements
End users must implement the following requirements when using OpenAI for the following high-risk use cases:
High-risk use case | Description | Mitigation |
---|---|---|
Applications involving chat or conversations | Applications that enable users to interact with a conversational agent | Verify that such applications are grounded or topical meaning that users interacting with the application do not have unrestricted access to query the model with general inputs that generate unrestricted outputs. |
Applications accessible to users outside your organization | Authenticate or monitor such users through one of the following mechanisms:
| |
Applications involving code generation or transformation scenarios |
| |
Applications enabling image inputs | You are required to ensure that such inputs are restricted to low risk and topical images. |