{ "@context": "https://schema.org", "@type": "SoftwareApplication", "@id": "https://www.insightits.com/products/prismguard.html", "slug": "prismguard", "canonicalName": "PrismGuard", "name": "PrismGuard — Audited Prompt-Injection Firewall", "canonicalUrl": "https://www.insightits.com/products/prismguard.html", "knowledgeUrl": "https://www.insightits.com/catalog/prismguard.json", "markdownUrl": "https://www.insightits.com/catalog/prismguard.md", "family": "aiSecurity", "infraFamily": "AI Security", "recordType": "product", "status": "published", "parentProduct": null, "whatItIs": "Self-hosted prompt-injection firewall and ChorusGraph security plug-in. Returns auditable allow/block with resolution_gate and decision_source. Opt-in profiles: web_chat, light/heavy ONNX, and 0.1.10 domain_pilot after train-first.", "problem": "Opaque guards and scanner-only stacks make it hard to audit which layer blocked a prompt, and cold holdout / domain traffic still needs a local, fail-closed decision.", "isNot": [ "Enterprise certification or SOC2", "A broad “beats LLM Guard on detection” claim", "A performance claim against Check Point AI Guardrails or NVIDIA NeMo", "Healthcare/finance domain packs with accuracy guarantees (use domain_pilot + train on your traffic)" ], "alternatives": [ { "name": "LLM Guard (PromptInjection scanner)", "relationship": "Published law-domain holdout compares a configured PrismGuard pipeline to one LLM Guard PromptInjection scanner configuration. Feature-asymmetric; not a whole-product guarantee.", "url": "https://www.insightits.com/compare/prismguard-vs-llm-guard.html" } ], "features": [ { "name": "Auditable Allow/Block", "description": "Every decision exposes resolution_gate and decision_source for compliance logs — not a black-box score." }, { "name": "Self-Hosted Firewall", "description": "Sits in front of the LLM (and can scan assistant output). Local ONNX classifier; no OpenAI required by default." }, { "name": "ChorusGraph Security Plug-in", "description": "make_guard_handler before RAG/LLM. Domain-agnostic for legal, healthcare, and finance — align ONNX with shadow + feedback on your traffic (law pack is published proof)." } ], "architecture": "Layered local rules + taxonomy + optional ONNX + optional judge escalation. Every decision reports resolution_gate and decision_source. ChorusGraph wiring: make_guard_handler + route_after_guard before cache/RAG/LLM.", "useCases": [ "Standalone prompt screening before RAG or LLM", "ChorusGraph security port for legal, healthcare, and finance traffic", "Train-first domain_pilot after you have labeled traffic" ], "benchmarks": { "published": true, "summary": "In the published 14-attack law holdout, the configured PrismGuard pipeline blocked 14/14 while the measured LLM Guard scanner blocked 9/14; both allowed 25/25 normal holdout cases. Separately, finance mid bake-off shows Prism pack PI 100% vs peer 85% vs AgentCore 45% (task ties AgentCore at 100%).", "disclosures": [ "Domain-aligned integrated firewall pipeline versus one LLM Guard prompt-injection scanner configuration (law). Separate finance mid bake-off versus LangGraph+LLM Guard and AgentCore Runtime.", "PrismGuard uses its law overlay, authored seed, ONNX artifact, and selective escalation on the law holdout. Finance mid uses domain-calibrated finance artifact + pack wiring. Results are not model-only or whole-product guarantees.", "Broadly beats LLM Guard on detection", "Universal prompt-injection accuracy", "Beat AgentCore / AWS overall", "Cross-host latency win vs AgentCore", "Mean LLM efficiency win vs AgentCore", "Enterprise healthcare or finance certification", "Performance superiority over Lakera Guard or NVIDIA NeMo Guardrails", "Enterprise-ready without a design-partner pilot" ], "urls": [ "https://github.com/insightitsGit/PrismGuard/blob/master/benchmark/law/results/current/COMPARISON_REPORT.md", "https://github.com/insightitsGit/PrismGuard/blob/master/benchmark/law/results/current/COMPARISON_REPORT.md" ], "lawHoldout": { "attackHoldout": { "n": 14, "prismGuardBlocked": 14, "baselineBlocked": 9 }, "normalHoldout": { "n": 25, "prismGuardAllowed": 25, "baselineAllowed": 25 }, "requestLatencyMeanMs": { "prismGuard": 211.16, "baseline": 352.73 }, "judgeEscalationRate": 0.0701 }, "financeMid": { "id": "mid_pla_faqfix_20260724", "allowedClaim": "On FinancePackBench mid (n=100/lane, seed 42), the Prism pack on PrismGuard 0.1.10 blocked 100% of finance PI attacks with 100% benign allow and 100% task success — ahead of LangGraph+LLM Guard on PI (85%) and AWS AgentCore Runtime mid on PI (45%). AgentCore tied task at 100%. Same-host task P50 ~2.4s vs LangGraph ~3.7s. Mean LLM 0.60 vs LangGraph 1.11 is a fair efficiency cite; do not compare mean LLM to AgentCore.", "notSupported": [ "Beat AWS / AgentCore overall", "Cross-host latency win vs AgentCore", "Mean LLM / efficiency win vs AgentCore (stubbed counter)", "Production RAG grounding quality (planted suite ~30%)", "100% PI forever / zero-day proof (holdout n=20 attacks)", "Enterprise finance-pack certification without customer retrain" ], "lanes": { "P": { "taskPct": 100, "piAttackPct": 100, "benignPct": 100, "groundPct": 30, "meanLlm": 0.6, "taskP50Ms": 2401, "llmMeter": "real ChorusGraph usage" }, "L": { "taskPct": 95, "piAttackPct": 85, "benignPct": 100, "groundPct": 30, "meanLlm": 1.11, "taskP50Ms": 3739, "llmMeter": "real LangGraph usage" }, "A": { "taskPct": 100, "piAttackPct": 45, "benignPct": 100, "groundPct": 26.7, "meanLlm": 0.4, "taskP50Ms": 4042, "llmMeter": "stub — hardcoded 1 per task; not comparable" } } } }, "github": "https://github.com/insightitsGit/PrismGuard", "pypi": "https://pypi.org/project/prismguard/0.1.10/", "documentation": [ { "title": "Prompt-injection firewall guide", "url": "https://www.insightits.com/guides/prompt-injection-firewall.html" }, { "title": "PrismGuard vs LLM Guard", "url": "https://www.insightits.com/compare/prismguard-vs-llm-guard.html" }, { "title": "LLM security guardrails guide", "url": "https://www.insightits.com/compare/llm-security-guardrails.html" }, { "title": "Design notes", "url": "https://github.com/insightitsGit/PrismGuard/blob/master/docs/prismguard-design.md" }, { "title": "Integration guide", "url": "https://github.com/insightitsGit/PrismGuard/blob/master/docs/integration-guide.md" }, { "title": "https://github.com/insightitsGit/PrismGuard/blob/master/benchmark/law/results/current/COMPARISON_REPORT.md", "url": "https://github.com/insightitsGit/PrismGuard/blob/master/benchmark/law/results/current/COMPARISON_REPORT.md" }, { "title": "enterprise", "url": "https://github.com/insightitsGit/PrismGuard/blob/master/docs/enterprise-product-model.md" }, { "title": "userUpdates", "url": "https://github.com/insightitsGit/PrismGuard/blob/master/docs/user-updates.md" }, { "title": "ci", "url": "https://github.com/insightitsGit/PrismGuard/actions/workflows/ci.yml" }, { "title": "onnxRelease", "url": "https://github.com/insightitsGit/PrismGuard/releases/tag/v0.1.2" }, { "title": "chorusgraphGithub", "url": "https://github.com/insightitsGit/ChorusGraph" } ], "install": "pip install \"prismguard[prism,guard-model]==0.1.10\". Apache-2.0 open core — free forever. Optional ops: ChorusControl Enterprise (CONTROL).", "version": "0.1.10", "relatedProducts": [ "chorusgraph", "prismshine", "prism-pack" ], "publisher": { "@type": "Organization", "name": "Insight IT Solutions", "legalName": "Insight IT Solutions LLC", "url": "https://www.insightits.com" } }