Custom Content Policies
Overview
Content policies can be used to create customized policies that are aligned to specific organizational requirements. Examples of custom policies include: avoiding financial advice, not mentioning a particular competitor. Input Content Policies can be used to detect non-compliant user inputs, while Output Content Policies can be used to detect non-compliant model responses
Content Policy Actions
Content policies currently enable flagging and blocking content.
- Flag: allow user inputs and model outputs containing toxic content, but flag input or output in moderator view
- Block: block user input or model output containing toxic content
Out-of-the-box Policy Inventory
In addition to providing tooling for custom guardrail creation, Dynamo Guard provides the following default guardrails to help your enterprise address common model safety and compliance scenarios.
| Policy | Input or Output | Definition | Date Updated |
|---|---|---|---|
| Prompt Injection | Input | Detects prompt injection attacks. | 07-15-2024 |
| Legal Advice | Input | Detects user inputs requesting legal advice. | 07-15-2024 |
| Financial Advice | Input | Detects user inputs requesting financial or investment advice. | 07-15-2024 |
| Prohibit Discrimination (Coming Soon) | Input | Prohibits prompts that discriminate or are discriminatory in nature towards any individual or group of individuals. | Coming Soon |
| Material Non-Public Information (Coming Soon) | Input | Prohibits prompts that include Material Non-Public Information. | Coming Soon |
| Compensation Data (Coming Soon) | Input | Prohibits prompts that request or provibe sensitive compensation data. | Coming Soon |