PyRIT
PyRIT copied to clipboard
Azure
The Python Risk Identification Tool for generative AI (PyRIT) is an open access automation framework to empower security professionals and machine learning engineers to proactively find risks in their...
#### Describe the issue linked to the documentation Hello! While running the tests for PyRIT locally with `make test-cov-xml`, I came across a few errors that resulted from aria2, pyodbc,...
#### Describe the issue linked to the documentation We have people running into issues with DuckDB semi-frequently where the solution is essentially to delete the results folder. We should have...
## Is your feature request related to a problem? Please describe. Garak and CyberSec have insecure code generation detectors. As I understand it, that means they have a scorer LLM...
TLDR; We want help creating issues where Garak, Giskard, or CyberSecEval support an attack scenario where PyRIT doesn't support it. #### Background Other tools like Garak, Giskard, and CyberSecEval have...
## Description Adding evaluation datasets for likert scale datasets + scoring scales used + demo in notebook
Going through GitHub issues, we've had so many community contributions related to datasets, and we have so many remaining open issues. It's a really easy way to add value to...
FIX: Issue 1204 - Relax Docker requirements.txt constraints to resolve dependency conflict ## Description Fixes dependency conflict in Docker build that prevented container creation. The issue was caused by overly...
## Description Migrating (back) to the openai SDK since it appears to have matured to a point where we can actually use it. This simplifies some things for us as...
#### Describe the bug Docker container build fails due to conflicting dependencies between `notebook==7.3.2` and `jupyterlab==4.4.8`. The `notebook 7.3.2` package requires `jupyterlab=4.3.4`, but pip attempts to install `jupyterlab==4.4.8`, causing a...
Addition of specific health dataset containing disinformation topics with jailbreaks. **Description** - Jailbreak dataset from https://doi.org/10.1136/bmj-2023-078538 - Two techniques: - Six prompts each targeting health topics - Four demographics for...
