{"full_name":"microsoft/PyRIT","owner":"microsoft","name":"PyRIT","shard":"1","url":"https://github.com/microsoft/PyRIT","description":"Open-source AI red teaming toolkit from Microsoft. Test generative AI systems with automated scans, adaptive multi-turn attacks, and a web UI.","homepage":"https://microsoft.github.io/PyRIT/","topics":["ai-red-team","generative-ai","red-team-tools","responsible-ai","ai-attacks","ai-evals","ai-evaluation","ai-evaluation-tools","ai-red-team-tool","ai-red-teaming","ai-safety","ai-safety-research","ai-security","ai-security-testing","ai-security-tool"],"tag_pages":["ai-security","ai-security-tool","generative-ai","responsible-ai"],"language":"Python","license":"MIT","avatar":"https://avatars.githubusercontent.com/u/6154722?v=4","created_at":"2023-12-12T15:46:28Z","pushed_at":"2026-11-16T02:07:19Z","metrics":{"stars":4871,"forks":1063,"watchers":53,"issues":846,"prs":3450,"releases":25,"issues_open":131,"issues_closed":715,"prs_open":146,"prs_closed":375,"prs_merged":2929,"discussions":12},"growth":{"day":{"stars":22,"forks":5,"base_day":"2026-11-15"}},"history":{"dates":["2026-11-15","2026-11-16"],"stars":[4849,4871],"forks":[1058,1063],"watchers":[53,53],"issues_open":[139,131],"prs_open":[144,146]},"boards":{},"same_owner":[{"full_name":"microsoft/markitdown","name":"markitdown","shard":"1","boards":{"daily":151,"weekly":176,"monthly":103,"total":36}},{"full_name":"microsoft/vscode","name":"vscode","shard":"1","boards":{"daily_watchers":21,"total":37}},{"full_name":"microsoft/PowerToys","name":"PowerToys","shard":"1","boards":{"daily_watchers":111,"total":81}},{"full_name":"microsoft/generative-ai-for-beginners","name":"generative-ai-for-beginners","shard":"1","boards":{"daily_forks":40,"total":98}},{"full_name":"microsoft/TypeScript","name":"TypeScript","shard":"1","boards":{"daily_watchers":120,"total":120}},{"full_name":"microsoft/terminal","name":"terminal","shard":"1","boards":{"total":133}},{"full_name":"microsoft/playwright","name":"playwright","shard":"1","boards":{"total":143}},{"full_name":"microsoft/Web-Dev-For-Beginners","name":"Web-Dev-For-Beginners","shard":"1","boards":{"total":155}},{"full_name":"microsoft/ML-For-Beginners","name":"ML-For-Beginners","shard":"1","boards":{"total":173}},{"full_name":"microsoft/AI-For-Beginners","name":"AI-For-Beginners","shard":"1","boards":{}},{"full_name":"microsoft/AirSim","name":"AirSim","shard":"1","boards":{}},{"full_name":"microsoft/BitNet","name":"BitNet","shard":"1","boards":{}},{"full_name":"microsoft/Bringing-Old-Photos-Back-to-Life","name":"Bringing-Old-Photos-Back-to-Life","shard":"1","boards":{}},{"full_name":"microsoft/CNTK","name":"CNTK","shard":"1","boards":{}},{"full_name":"microsoft/Data-Science-For-Beginners","name":"Data-Science-For-Beginners","shard":"1","boards":{}},{"full_name":"microsoft/FluidFramework","name":"FluidFramework","shard":"1","boards":{}},{"full_name":"microsoft/IoT-For-Beginners","name":"IoT-For-Beginners","shard":"1","boards":{}},{"full_name":"microsoft/MMdnn","name":"MMdnn","shard":"1","boards":{}},{"full_name":"microsoft/Ontology-Playground","name":"Ontology-Playground","shard":"1","boards":{}},{"full_name":"microsoft/Procexp-Mac","name":"Procexp-Mac","shard":"1","boards":{"daily_rate":96,"weekly_rate":32,"monthly_rate":87}}],"issues":[{"number":3166,"title":"Python 3.11 adoption opportunities: claim one improvement at a time","comments":205,"created_at":"2026-10-13T05:03:12Z"},{"number":266,"title":"FEAT Add adaptive jailbreaking","comments":13,"created_at":"2024-06-28T20:15:14Z"},{"number":2241,"title":"FEAT: MaliciousToolCallInjection attack strategy for agentic pipeline red teaming","comments":12,"created_at":"2026-07-21T21:52:35Z"},{"number":4216,"title":"MAINT: Reduce the keyword-only baseline, starting with private helpers","comments":11,"created_at":"2026-11-11T20:13:27Z"},{"number":2475,"title":"Labelled responses from remote datasets can't reach `HumanLabeledDataset`","comments":11,"created_at":"2026-08-24T13:12:28Z"},{"number":791,"title":"DOC Adversarial model doc page","comments":11,"created_at":"2025-03-14T23:43:54Z"},{"number":2489,"title":"FEAT Add generic GCG model adapters and an open-weight SLM compatibility matrix","comments":10,"created_at":"2026-08-25T19:17:18Z"},{"number":3579,"title":"FEAT: Report confidence intervals and scorer specificity alongside attack success rates in analytics","comments":8,"created_at":"2026-10-27T01:39:28Z"}],"discussions":[{"number":586,"title":"Using LM Studio","comments":0,"category":"General","created_at":"2024-12-03T09:30:34Z"},{"number":370,"title":"Finetuning OpenAI model with prompt injections, to act as the attacker LLM","comments":0,"category":"General","created_at":"2024-09-12T08:32:42Z"},{"number":348,"title":"Tutorial on PyRIT on AWS","comments":1,"category":"General","created_at":"2024-08-23T23:13:04Z"},{"number":346,"title":"Can we use PyRIT on local HF models?","comments":1,"category":"General","created_at":"2024-08-23T22:20:37Z"},{"number":326,"title":"A little confused on merged code (such as EndTokenRedTeamingOrchestrator)","comments":1,"category":"Q&A","created_at":"2024-08-14T19:00:29Z"},{"number":311,"title":"pyrit with OpenAI key","comments":1,"category":"Q&A","created_at":"2024-08-01T15:19:10Z"},{"number":295,"title":"Support Custom API endpoint","comments":2,"category":"Ideas","created_at":"2024-07-24T11:41:03Z"},{"number":250,"title":"When is the next release?","comments":1,"category":"Q&A","created_at":"2024-06-20T17:29:30Z"}],"latest_release":{"name":"v1.2.0","tag":"v1.2.0","published_at":"2026-11-13T19:28:10Z","body":"# PyRIT v1.2.0\r\n\r\n## What's changed?\r\n\r\nPyRIT `v1.2.0` delivers a major upgrade to the CoPyRIT GUI, making it easier to launch, monitor, compare, and report on security assessments. Reusable scenario configurations, grouped launches, persistent engagements and findings, and richer result analytics turn individual scans into repeatable, reviewable workflows. The release also expands attack techniques, targets, converters, datasets, and scoring options.\r\n\r\nPython support is now `>=3.11, <3.15`.\r\n\r\n### 🌟 Highlights\r\n\r\n- **A major CoPyRIT GUI update** brings scan launch, live progress, history, scoring, comparison, and review into one workflow.\r\n- **Scenario presets and suites** make assessments reusable and allow related scenario configurations to be launched together.\r\n- **Operations and Findings** organize engagements, capture human-reviewed issues and evidence, and support reporting.\r\n- **New attacks and targets** expand coverage with GOAT, role-play and jailbreak techniques, A2A, HackAPrompt, and GitHub Copilot.\r\n- **More scoring and benchmark options** add ASSERT, Microsoft decision-model scoring, local scorers, and several major datasets.\r\n- **Improved reliability and security** strengthen execution cleanup, resume behavior, validation, authorization, and persistence.\r\n\r\n### 🤝 Community contributions\r\n\r\nCommunity contributors drove a substantial part of this release: **108 external contributors authored 507 merged pull requests—58% of the 882 release pull requests identified from the release cut.** Thank you to everyone who contributed features, datasets, tests, documentation, bug fixes, and reviews to PyRIT `v1.2.0`. See the [New Contributors](#new-contributors) section below.\r\n\r\n### ⚠️ Breaking changes\r\n\r\n- **Python 3.10 is no longer supported.** Upgrade to Python 3.11–3.14 before installing PyRIT 1.2.0.\r\n- **Backend and client artifacts must match exactly.** Compatibility-enforcing CoPyRIT backends reject CLI or browser clients built from a different source commit, even when the package version matches. Deploy the backend, frontend, and CLI from the same source commit, and do not roll back below the first compatibility-enforcing backend while newer clients remain active ([#2829](https://github.com/microsoft/PyRIT/pull/2829)).\r\n- **Scenario construction and dataset configuration changed.** Custom scenarios must use the current parameter/context APIs and argument-free `initialize_async()`. Dataset selection now uses explicit sources, fetch policies, and limits; use `\"all\"` rather than `None` for unlimited selection ([#3404](https://github.com/microsoft/PyRIT/pull/3404), [#2956](https://github.com/microsoft/PyRIT/pull/2956), [#2849](https://github.com/microsoft/PyRIT/pull/2849), [#2904](https://github.com/microsoft/PyRIT/pull/2904), [#3143](https://github.com/microsoft/PyRIT/pull/3143)).\r\n- **Backend scenario reads are async-only.** Integrations using synchronous scenario-read methods must migrate to and await the asynchronous APIs ([#2953](https://github.com/microsoft/PyRIT/pull/2953)).\r\n- **Custom `MessageScorer` subclasses must use the expectation-aware scoring hook.** Replace overrides of `_score_piece_async(..., objective=...)` with `_score_piece_with_expectation_async(..., expectation=...)` and read scoring context from the expectation ([#3312](https://github.com/microsoft/PyRIT/pull/3312)).\r\n- **The Gandalf target and scorer were removed.** Remove imports and configuration references, and migrate to another supported or custom target/scorer ([#3467](https://github.com/microsoft/PyRIT/pull/3467)).\r\n- **Some API target creation now requires administrator access.** Non-admin users can no longer create targets that use server-stored credentials or execute model-repository code; use an administrator-created target or caller-provided credentials where supported ([#3247](https://github.com/microsoft/PyRIT/pull/3247), [#3173](https://github.com/microsoft/PyRIT/pull/3173)).\r\n- **API-created HTTP target uploads now use a serve","is_prerelease":false},"readme":true,"readme_zh":false,"news":[],"stale":false,"snapshot_at":"2026-11-16T09:23:14Z","history_start":"2026-09-06"}