September 3rd 2024

​

Prompt Injection Fine-Tuned Policy

We’ve refined our prompt injection policy to enhance performance with three sensitivity levels. This new approach allows you to select the sensitivity level that best suits your use case. The levels are defined as:

1. Level 1: Detects only clear cases of prompt injection. Ideal for minimizing false positives but might overlook ambiguous cases.
2. Level 2: Balanced detection. Effectively identifies clear prompt injections while reasonably handling ambiguous cases.
3. Level 3: Detects most prompt injections, including ambiguous ones.

​

PII Masking - New PII Policy Action

We’ve introduced a new action for our PII policy; PII masking, that replaces sensitive data with corresponding tags before the message is processed or sent. This ensures that sensitive information remains protected while allowing conversations to continue.

Example Before Masking:

Please send the report to john.doe@example.com and call me at 123-456-7890.

Example After Masking:

Please send the report to <EMAIL> and call me at <PHONE_NUMBER>.

​

API Keys Management

We’ve added a new API Keys table under the “My Account” section to give you better control over your API keys. You can now create and revoke API keys. For security reasons, you won’t be able to view the key again after creation, so if you lose this secret key, you’ll need to create a new one.

​

Navigation Between Dashboard and Projects

General Dashboard: You can now easily navigate from the general dashboard to your projects by simply clicking on any project in the active project section.

Project Dashboard: Clicking on any action or policy will take you directly to the project’s Session Explorer, pre-filtered by the same policy/action and date range. Additionally, “Clicking on the prompt/response graphs in the analytics report will also navigate you to the Session Explorer, filtered by the corresponding date range.

​

Policy Example Demonstrations

We’ve enhanced the examples section for each policy to provide clearer explanations. You can now view a sample conversation between a user and an LLM when a violation is detected and action is taken by Aporia. Simply click on “Examples” before adding a policy to your project to see which violations each policy is designed to prevent.

​

Improved Policy Configuration Editing

We’ve streamlined the process of editing custom policy configurations. Now, when you click “Edit Configuration”, you’ll be taken directly to the policy configuration page in the policy catalog. Once there, you can easily return to your project with the new “Back to Project” arrow.