Technology · Agentic → Guardrails · Open Source · wiki:deep

LLM Guard

Taxonomy

LLM Guard (Protect AI) is a security toolkit for LLM interactions: sanitization, harmful-language detection, data-leakage prevention, and prompt-injection resistance via input and output scanners (anonymize, secrets, toxicity, prompt injection, malicious URLs, and many more). Install with pip install llm-guard (Python ≥ 3.9); deployable as a library or API.

Primary README warning: the project has been archived — the repo and associated Hugging Face models are no longer actively developed or maintained. Treat as historical reference and pin commits if still used.

RepoDocsStack: agentic-guardrails

Why it matters here

Guardrails research still needs a catalog of I/O scanners for prompt injection, secrets, and toxicity. LLM Guard remains a useful shape reference for scanner libraries even though it is archived. Prefer Guardrails AI or NeMo Guardrails for maintained programmable rails; prefer Headroom for action-level HITL, not text scanners.

How it works

Wrap LLM calls with prompt scanners on the way in and output scanners on the way out. Each scanner targets a risk class (secrets, injection, toxicity, …). Optional API deployment fronts the same scanners for services.

  1. pip install llm-guard.
  2. Configure input scanners on user/tool text.
  3. Call the LLM.
  4. Run output scanners before acting on the response.

Related: guardrails-ai · headroom · nemo-guardrails · topics/13-governance-policy

Flow

Scroll inside the canvas to pan

When to reach for it

  • Use when: studying scanner-style I/O defenses or maintaining a pinned legacy integration.
  • Skip when: you need an actively maintained guard product (guardrails-ai, nemo-guardrails) or HITL action calibration (headroom).
  • Prefer instead: Guardrails Hub validators or NeMo rails for new installs.

Limits

  • Archived: no active maintenance — security scanners age poorly; do not pick for greenfield.
  • Model/HF coupling: some scanners depended on Protect AI HF models now frozen.
  • Not a sandbox / not HITL: text filters ≠ code isolation or approval budgets.
  • Not the ledger.

Linked from

Catalog backlinks — what points here (wiki “what links here”).

Features

Stacks

Peers

Document management

What we checked

Claims below are backed by science sources on disk.

Toolkit / scanner framing

protectai/llm-guard · MODERATE

“LLM Guard is a comprehensive tool designed to fortify the security of Large Language Models (LLMs).”

Docs / scanner list

LLM Guard documentation · MODERATE

“LLM Guard by Protect AI is a comprehensive tool designed to fortify the security of Large Language Models (LLMs).”

Peer contrast

NeMo Guardrails contrast · MODERATE

“THIS PROJECT HAS BEEN ARCHIVED.”

In this research stack

Features and peers linked from the catalog map — not a second product surface.

Features it supports

Same cell

Stack

Document management

Source

Primary repo github.com/protectai/llm-guard · Open Source

technologies/llm-guard/README.md

Research inventory

9 tags · 37 out · 38 in · 3 artifacts · 0 gaps · 34 corpus docs

Catalog tags

landscape.layer
Agentic
landscape.subcategory
Guardrails
license_tag
Open Source
maps.dm
present
maps.features
10
one_liner
Agentic
review.depth
science
slug
llm-guard
title
LLM Guard

Artifacts

  • dm_map · present · technologies/llm-guard/document-management.md
  • features_map · present · technologies/llm-guard/features.md
  • readme · present · technologies/llm-guard/README.md

Out · alternative_to

Out · dm_axis

Out · maps_to

In · alternative_to

In · dm_axis

In · in_stack

In · maps_to

Corpus tags

category
Agentic → Guardrails
dedication
open-source
feature
access-vs-authority
agent-supply-chain
autonomy-levels
emergency-control
human-oversight-guardrails
job-security
r2-13-governance-policy
r2-16-approval-calibration
security-identity-governance
wt-human-policy
kind
map_edge
tech_features
tech_quote
tech_readme
tech_science_source
tech_section
needs_deepen
false
quality
ok
section
Document management map
Evidence
Features map
GitHub map
How it works
Limits & failure modes
Links
Scientific notes
What it is
When to use / skip
Why it matters here
slug
llm-guard
source_id
llmguard-docs
llmguard-github
nemo-contrast
technology
llm-guard

Corpus documents (34)

map_edge · 10

  • llm-guard → access-vs-authority
  • llm-guard → agent-supply-chain
  • llm-guard → autonomy-levels
  • llm-guard → emergency-control
  • llm-guard → human-oversight-guardrails
  • llm-guard → job-security
  • llm-guard → r2-13-governance-policy
  • llm-guard → r2-16-approval-calibration
  • … +2 more

tech_features · 1

  • LLM Guard · features

tech_quote · 8

  • LLM Guard · llmguard-docs
  • LLM Guard · llmguard-docs
  • LLM Guard · llmguard-docs
  • LLM Guard · llmguard-github
  • LLM Guard · llmguard-github
  • LLM Guard · llmguard-github
  • LLM Guard · nemo-contrast
  • LLM Guard · nemo-contrast

tech_readme · 1

  • LLM Guard

tech_science_source · 3

  • LLM Guard · llmguard-docs
  • LLM Guard · llmguard-github
  • LLM Guard · nemo-contrast

tech_section · 11

  • LLM Guard · Document management map
  • LLM Guard · Evidence
  • LLM Guard · Features map
  • LLM Guard · GitHub map
  • LLM Guard · How it works
  • LLM Guard · Limits & failure modes
  • LLM Guard · Links
  • LLM Guard · Scientific notes
  • … +3 more