llm-guard
3,221The Security Toolkit for LLM Interactions.
Upstream archived. The repository of llm-guard is read-only: it gets no further changes. Upstream is archived (checked on GitHub 2026-11-21; last push 2026-07-08). The repository README says the project and its models on Hugging Face "are no longer under active development or maintained". Status set to deprecated; no successor is named.
Activity band details
Activity band A to F: a bin of a score that adds GitHub stars (log scale), a push to the upstream repository in the 12 months before `as_of`, the number of declared integration surfaces and, where the catalogue records them, citing papers and benchmark evaluations. It measures popularity and activity. It is not a level of evidence: no band means a paper cites the tool or that it was measured. Recency is judged as of 2026-11-22.
Computed band: F
(raw score 3.9).
Shown band: F
.
Facets:
documented_in=0evaluated_on=0integration_surfaces=0.5paper_citations=0recent_push=2stars=1.4
capped from E to F by maturity_status='deprecated'
Source of truth: pipeline/maturity/score.py
Description
ProtectAI's LLM Guard is a runtime security layer for LLM interactions. Pluggable input/output scanners (PII, secrets, prompt injection, jailbreak detection, profanity, toxicity, competitor detection) can be composed as a gate in front of LLM calls, with configurable allow/deny policies.
Why it's included
Canonical runtime-guard toolkit: covers both directions (user→LLM and LLM→user). Differentiated from nemo-guardrails by being library-first (no DSL).
Addresses failure modes
These tags are the catalog's claim about what llm-guard is built to catch or prevent. A tag is not a test result. SAICA-Bench has not measured llm-guard, so none of these tags is tested here.
- Security vulnerability — Insecure patterns, injection flaws, weak cryptography, secret leakage.
- Fabrication — Generating reference to non-existent APIs, packages, or symbols.