User Guide
- Manage Prompts
- Create Prompt
- Edit Version
- Generate Version
- Rename Version
- Delete Version
- Set Primary Version
- Compare Versions
- Token Count and Cost
- Update Prompt Use Case
- Copy Prompt to Workspace
- Move Prompt to Workspace
- Optimize Prompts
- Prompt Scoring
- Scoring Criteria
- Prompt Risk
- Prompt Balance
- Prompt Heatmap
- Connect AI Providers
- AI21 Labs
- Amazon Bedrock
- Anthropic
- Azure OpenAI
- Azure Serverless
- Cohere
- Deep Infra
- Fireworks AI
- Google Gemini
- Groq
- Hugging Face
- Hyperbolic
- Lambda
- Meta
- Mistral
- Moonshot AI
- OpenAI
- Perplexity
- Together AI
- xAI
Prompt Risk
On this page you'll find the following:
Get Started
Use the prompt risk tool to identify any possible bias, ethical concerns, or safety issues. This is especially useful for system prompts that could be the source of improper responses. This helps eliminate AI responses that could harm your company's brand and reputation.
To get started, go to the version's Risk tab and click the purple Analyze risk button.
This will bring up the Analyzing Risk modal box. After a couple seconds the modal box will go away and the risk results will be displayed.
The results show the risk level for potential bias, ethical concerns, safety issues, and the overall risk level. The right sidebar contains a summary of the analysis as well as how many tokens were used and how long it took to analyze the risk.
If a category is Medium or High risk, the offending prompt part will be listed under its corresponding category with the Issue as to why it's problematic and a Suggested Revision for how to fix it.
Risk Levels
Each category is assigned one of the following levels: Low, Medium, or High. The overall risk level is determined as such:
- If any of the categories is High risk, the overall risk level will be High.
- If there are no High risk categories but there is at least one Medium risk category, the overall risk level will be Medium.
- If all categories are Low risk, the overall risk level will be Low.
Auto-Improve Risk
For prompts with Medium or High risk, a purple Improve button will appear under the summary in the right sidebar. Clicking this button will replace any offending prompt parts with their suggested revision and create a new prompt version with those changes in place.
The Improve button does not display for prompts with Low overall risk.
When the prompt improvement is complete, you'll be taken to the prompt Compare screen so you can see the changes that were made side-by-side with the previous version
Analyze Risk with Different Models
To change which AI model is used to analyze the prompt risk, click the Settings link near the top-right of the screen.
This will open a side panel where you can choose different models to use for the risk analysis. Make your model selection, click the green Apply button, then run the risk analysis.
