Data as of Sep 26, 2026 · Based on 4,029,442 AI responses across 13,338 prompts · See how Parse measures this
Rebuffgithub.comRebuff is a framework that protects AI applications from prompt injection attacks using a multi-layer defense, including heuristics, LLM-based detection, a vector database of embeddings, and canary tokens. It provides JavaScript/TypeScript and Python SDKs and can be installed via pip, with examples for detecting prompt injections and canary word leakage. The project is described as a prototype and does not guarantee 100% protection.
The market map · 5 of 100 labelled
LLM Firewall and Guardrail Platforms →Where Rebuff ranks in AI
Lakera GuardPicked instead5 questions
NVIDIA NeMo GuardrailsPicked instead1 question
Palo Alto NetworksRanking↑5 to #12
Prompt SecurityRanking↑1 to #6
Guardrails AIRanking#7Excerpts where Rebuff appeared in the AI's answer
Rebuff: A self-hardening prompt injection detection gateway that uses a multi-layered defense approach (heuristics, LLM-based checks, and canary tokens) to catch prompt injections.
Rebuff : An open-source security layer designed specifically to detect and deflect prompt injection attacks using heuristics, vector databases, and canary tokens.
Excerpts where Rebuff appeared in the AI's answer
Rebuff / Other Open Source Frameworks: Techniques such as "tool output sanitizers" act as a judge model between an AI agent and its environment, effectively mitigating indirect prompt injections.
Rebuff: A specialized framework designed to detect prompt injection attacks, often used for input verification.
Excerpts where Rebuff appeared in the AI's answer
Rebuff.ai : A self-hostable/open-source security framework designed to protect LLM applications against prompt injection.
Excerpts where Rebuff appeared in the AI's answer
Rebuff.ai: A specialized tool designed to detect prompt injection attacks and strengthen LLM applications against prompt leakage.
Rebuff: A specialized tool focused on detecting and mitigating prompt injection attacks at runtime
Excerpts where Rebuff appeared in the AI's answer
Rebuff / HiddenLayer : Specializing in runtime security for agentic systems, safeguarding multi-step autonomous tool-use from indirect prompt injection.
