LLM - Blue Team
Wprowadzenie
Ta strona została stworzona dla członków niebieskiej drużyny, badaczy sztucznej inteligencji oraz entuzjastów defensywnej strony bezpieczeństwa LLM.
Znajdziesz tu narzędzia, zasoby i uwagi dotyczące ochrony dużych modeli językowych (LLM).
Nadziędzia
Open Source
| Date |
Repo |
Description |
Stars |
Watchers |
Link |
| 2026-04-03 |
guardrails |
Adding guardrails to large language models. |
⭐ 6630 |
👁️ 36 |
guardrails |
| 2026-04-02 |
Guardrails |
NeMo Guardrails is an open-source toolkit for easily adding programmable guardrails to LLM-based conversational systems. |
⭐ 5920 |
👁️ 40 |
Guardrails |
| 2026-03-31 |
PurpleLlama |
Set of tools to assess and improve LLM security. |
⭐ 4102 |
👁️ 68 |
PurpleLlama |
| 2026-03-27 |
openguardrails |
Protect every action your agent takes. |
⭐ 327 |
👁️ 4 |
openguardrails |
| 2025-12-11 |
alibi-detect |
Algorithms for outlier, adversarial and drift detection |
⭐ 2511 |
👁️ 35 |
alibi-detect |
| 2025-10-08 |
granite-guardian |
The Granite Guardian models are designed to detect risks in prompts and responses. |
⭐ 136 |
👁️ 9 |
granite-guardian |
| 2025-09-03 |
llm-guard |
The Security Toolkit for LLM Interactions |
⭐ 2776 |
👁️ 26 |
llm-guard |
| 2025-03-22 |
llm-defense |
An easy-to-use Python framework to defend against jailbreak prompts. |
⭐ 21 |
👁️ 1 |
llm-defense |
| 2024-11-22 |
langkit |
🔍 LangKit: An open-source toolkit for monitoring Large Language Models (LLMs). 📚 Extracts signals from prompts & responses, ensuring safety & security. 🛡️ Features include text quality, relevance metrics, & sentiment analysis. 📊 A comprehensive tool for LLM observability. 👀 |
⭐ 980 |
👁️ 15 |
langkit |
| 2024-11-22 |
langkit |
🔍 LangKit: An open-source toolkit for monitoring Large Language Models (LLMs). 📚 Extracts signals from prompts & responses, ensuring safety & security. 🛡️ Features include text quality, relevance metrics, & sentiment analysis. 📊 A comprehensive tool for LLM observability. 👀 |
⭐ 980 |
👁️ 15 |
langkit |
| 2024-07-16 |
HeimdaLLM |
Constrain LLM output |
⭐ 113 |
👁️ 3 |
HeimdaLLM |
| 2024-01-31 |
vigil-llm |
⚡ Vigil ⚡ Detect prompt injections, jailbreaks, and other potentially risky Large Language Model (LLM) inputs |
⭐ 470 |
👁️ 9 |
vigil-llm |
| 2024-01-31 |
vigil-llm |
⚡ Vigil ⚡ Detect prompt injections, jailbreaks, and other potentially risky Large Language Model (LLM) inputs |
⭐ 470 |
👁️ 9 |
vigil-llm |
| 2024-01-25 |
rebuff |
LLM Prompt Injection Detector |
⭐ 1455 |
👁️ 15 |
rebuff |
| 2024-01-25 |
rebuff |
LLM Prompt Injection Detector |
⭐ 1455 |
👁️ 15 |
rebuff |
Komercyjne
| Firma |
Narzędzie |
Model komercyjny |
Kraj powstania |
Kraj gł. udziałowca |
Link |
| Lakera |
Lakera Guard |
Usage-based API. Najpopularniejszy komercyjny firewall runtime (In-line) dla LLM. |
Szwajcaria |
Izrael (Check Point) |
Link |
| Cisco |
Cisco AI Defense |
Enterprise Subscription. Ochrona na poziomie warstwy sieciowej zintegrowana z infrastrukturą. |
USA |
USA |
Link |
| Coralogix |
Coralogix AI |
SaaS Enterprise. Monitoring (Observability) i guardrails. |
Izrael |
USA / Izrael |
Link |
| HiddenLayer |
AISPM Platform |
Annual License. Specjalistyczna ochrona własności intelektualnej i integralności modeli. |
USA |
USA |
Link |
| Cato Networks |
Cato AI Security |
Enterprise License. Ochrona "Shadow AI" i agentów wbudowana w platformę SASE. |
Izrael |
Izrael / USA |
Link |
| DataSunrise |
Database AI Security |
Per Instance License. Bezpieczeństwo danych (maskowanie PII) w bazach danych zintegrowanych z AI. |
USA / Izrael |
USA |
Link |
| CrowdStrike |
Falcon AI Guard |
SaaS / Pay-as-you-go. Ochrona runtime i filtrowanie promptów. |
USA |
USA |
Link |
| Promptfoo |
Promptfoo Guard |
Seat-based / Enterprise. Komercyjne guardrails |
USA |
USA (Insight Partners) |
Link |
Ochrona prywatności danych w modelach LLM
| Date |
Repo |
Description |
Stars |
Watchers |
Link |
| 2026-03-25 |
LangBiTe |
A Bias Tester framework for LLMs |
⭐ 24 |
👁️ 3 |
LangBiTe |
| 2024-08-12 |
anonLLM |
anonLLM: Anonymize Personally Identifiable Information (PII) for Large Language Model APIs |
⭐ 77 |
👁️ 1 |
anonLLM |
Inspiracje i pomysły
Prompt Shields