PyRIT

license

MIT(OSI)

source

public(full Python framework)

self-host

yes

service

none

core features withheld

no

PyRIT is released under the MIT license, with its full source published and installable from PyPI or a local container. No paid tier, enterprise build or license key gates any part of the framework.

• https://github.com/microsoft/PyRIT recorded 2026-08-13

  Repo metadata records license spdxId - MIT. README describes PyRIT as 'an open source framework built to empower security professionals and engineers to proactively identify risks in generative AI systems' and points at the docs site for install and use; no hosted tier, no enterprise directory and no license key on the page.

• https://microsoft.github.io/PyRIT/1.1.0/ recorded 2026-09-29

  The PyRIT 1.1.0 documentation home describes "a flexible, extensible framework for assessing the security and safety of generative AI systems at scale", with "Run multi-turn attack strategies like Crescendo, TAP, and Skeleton Key against AI systems", "Run standardized evaluation scenarios at large scale", a local GUI started with "serves webapp on http://localhost:8000/", targets spanning "Test OpenAI, Azure, Anthropic, Google, HuggingFace, custom HTTP endpoints or WebSockets", and scoring with "true/false, Likert scale, classification, and custom scorers". Install is "pip install pyrit"; searched the body for enterprise, license key, pricing and subscription: none.

• https://microsoft.github.io/PyRIT/1.1.0/getting-started/install recorded 2026-09-29

  The PyRIT install page offers user installs where "User installations" come from the "latest stable release" on PyPI, and contributor routes that "Install from source with uv in editable mode" against the "latest development code" on the main branch, plus a "Pre-configured container with JupyterLab" for local use. Searched the body for enterprise, license key, pricing and subscription: none.

• https://pypi.org/project/pyrit/ recorded 2026-09-29

  The PyPI project page for pyrit 1.1.0 (released Sep 4, 2026, author "Microsoft AI Red Team") lists "License expression" as "MIT", offers a "Source distribution for pyrit 1.1.0" as pyrit-1.1.0.tar.gz alongside the wheel, and carries the project README describing "The Python Risk Identification Tool for generative AI (PyRIT) is an open source" framework. Searched the body for enterprise, license key and commercial: none; the only paid match is a PyPI sponsor link.

GitHub stars are the only adoption signal published for PyRIT: no download, install or customer figure exists for the framework. A star is not a use, and Microsoft's backing is provenance rather than evidence of adoption.

• https://api.github.com/repos/microsoft/PyRIT recorded 2026-08-12

  stargazers_count = 4,285 for microsoft/PyRIT

PyRIT automates multi-turn attacks such as Crescendo and TAP, runs standardized harm scenarios at scale, and scores responses with LLM-based, classifier or custom scorers. It works against targets from OpenAI, Azure, Anthropic and Google to Hugging Face models and custom HTTP endpoints, and adds a command-line scanner and a GUI for human-led red teaming.

• https://api.github.com/repos/microsoft/PyRIT/contents/pyrit recorded 2026-08-13

  Top-level package listing carries converter, datasets and prompt_normalizer (adversarial prompt generation), executor and scenario (attack-strategy orchestration), score (scoring), and prompt_target with backend and auth (multi-provider targets) - the three things this reading is based on, as shipped modules.

• https://github.com/microsoft/PyRIT recorded 2026-08-13

  README describes 'an open source framework built to empower security professionals and engineers to proactively identify risks in generative AI systems'; the component detail it used to carry now lives on the docs site.

• https://microsoft.github.io/PyRIT/1.1.0/ recorded 2026-09-29

  The PyRIT 1.1.0 documentation home describes "a flexible, extensible framework for assessing the security and safety of generative AI systems at scale", with "Run multi-turn attack strategies like Crescendo, TAP, and Skeleton Key against AI systems", "Run standardized evaluation scenarios at large scale", a local GUI started with "serves webapp on http://localhost:8000/", targets spanning "Test OpenAI, Azure, Anthropic, Google, HuggingFace, custom HTTP endpoints or WebSockets", and scoring with "true/false, Likert scale, classification, and custom scorers". Install is "pip install pyrit"; searched the body for enterprise, license key, pricing and subscription: none.