agencyenterprise/PromptInject

PromptInject is a framework that assembles prompts in a modular fashion to provide a quantitative analysis of the robustness of LLMs to adversarial prompt attacks. 🏆 Best Paper Awards @ NeurIPS ML Safety Workshop 2022

51
/ 100
Established

PromptInject helps you understand how robust your large language model (LLM) is against malicious user inputs that try to hijack its intended purpose or leak its internal instructions. It takes your LLM and various attack prompts as input, then shows you how easily the model can be misled. This tool is for AI safety researchers, machine learning engineers, and product managers responsible for deploying LLMs in real-world applications.

465 stars. No commits in the last 6 months. Available on PyPI.

Use this if you are deploying a customer-facing LLM and need to rigorously test its security and reliability against adversarial inputs.

Not ideal if you are looking for a general-purpose tool to improve LLM performance or fine-tune models for specific tasks.

AI-safety LLM-security adversarial-testing prompt-engineering model-robustness
Stale 6m
Maintenance 0 / 25
Adoption 10 / 25
Maturity 25 / 25
Community 16 / 25

How are scores calculated?

Stars

465

Forks

44

Language

Python

License

MIT

Last pushed

Feb 26, 2024

Commits (30d)

0

Dependencies

4

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/prompt-engineering/agencyenterprise/PromptInject"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.