# Arthur Shield vs Llama Guard

**Leader by vioscaleAI score:** Arthur Shield

| Attribute | Arthur Shield | Llama Guard |
|---|---|---|
| **vioscaleAI score** | 62.6 (54% (medium)) | 43.9 (25% (low)) |
| activity.commits_last_30d | - | 3 |
| adoption.dependent_repos | - | 0 |
| adoption.github_stars | - | 4,431 |
| content.faq | `[{"answer":"A platform that locates and provides continuous monitoring of AI agents deployed across cloud and on-premises environments, offering security controls, audit capabilities, behavioral analysis, and policy enforcement. It is indexed under AI Guardrails.","source":"https://www.arthur.ai","question":"What is Arthur Shield?","confidence":0.6},{"answer":"Arthur Shield does not publish its prices. Pricing is quoted on request, across 2 plans (Free and Enterprise), so the figure depends on your seat count and requirements. We record this as a pricing-transparency signal rather than guessing a number. Pricing changes often, so verify at source before relying on it.","source":"https://www.arthur.ai","question":"How much does Arthur Shield cost?","confidence":0.6},{"answer":"We have confirmed browser-based access to Arthur Shield. That is the extent of what we could verify from public sources, so it may well offer desktop or mobile clients we have not indexed.","source":"https://www.arthur.ai/pricing-2025","question":"What platforms does Arthur Shield support?","confidence":0.6},{"answer":"Yes. Arthur Shield can be deployed cloud / SaaS, on-premise and self-hosted, so it does not have to run on the vendor's infrastructure.","source":"https://www.arthur.ai","question":"Can Arthur Shield be self-hosted?","confidence":0.6},{"answer":"We have confirmed 6 integrations for Arthur Shield, including OpenTelemetry, Snowflake, AWS Bedrock, Google Vertex AI, MCP and Claude Code. This is what we could verify from public sources, so the vendor may support others we have not indexed.","source":"https://www.arthur.ai/pricing-2025","question":"What does Arthur Shield integrate with?","confidence":0.6},{"answer":"We have independently confirmed SOC 2 and GDPR for Arthur Shield. Certifications we do not list are ones we have not been able to verify from public sources, which is not the same as Arthur Shield not holding them. Always confirm compliance directly before you rely on it.","source":"https://www.arthur.ai/blog/eu-ai-act-compliance-for-ai-agents","question":"What security certifications does Arthur Shield have?","confidence":0.7},{"answer":"Arthur Shield is available worldwide. Its primary market is the United States.","source":"https://www.arthur.ai/pricing-2025","question":"Where is Arthur Shield available?","confidence":0.5}]` | `[{"answer":"A suite of input/output moderation models and security evaluation benchmarks designed to help developers build safer generative AI applications. Includes tools to detect prompt injection, jailbreaking attempts, and other violating content. It is indexed under AI Guardrails.","source":"https://github.com/pricing","question":"What is Llama Guard?","confidence":0.6},{"answer":"Llama Guard is open source, so it can be self-hosted and used at no licence cost. Pricing changes often, so verify at source before relying on it.","source":"https://github.com/pricing","question":"Is Llama Guard free to use?","confidence":0.6},{"answer":"Llama Guard supports a command-line interface. Platforms we have not confirmed are simply not listed here rather than ruled out.","source":"https://github.com/pricing","question":"What platforms does Llama Guard support?","confidence":0.6},{"answer":"Yes. Llama Guard can be deployed self-hosted, so it does not have to run on the vendor's infrastructure.","source":"https://github.com/pricing","question":"Can Llama Guard be self-hosted?","confidence":0.6}]` |
| deployment.options | `{"cloud":true,"on_prem":true,"self_hosted":true}` | `{"self_hosted":true}` |
| description.long | An open-source tool designed to enforce safety guardrails in AI/ML workloads, detect quality issues like data leakage and hallucinations in real-time, manage prompts across applications, and monitor model performance and RAG pipelines. | Open-source moderation models that detect and help prevent harmful content in AI language model inputs and outputs, with capabilities including prompt injection and jailbreak attack detection. |
| features.capabilities | `{"iso_27001":false,"soc2_type_ii":true,"pricing_model":"flat_enterprise_tier","architecture_model":"managed_cloud_inline_proxy","rag_retrieval_chunk_poisoning_defense":true,"prompt_injection_and_jailbreak_blocking":true,"inbound_pii_and_credit_card_masking_redaction":true,"streaming_token_interception_and_chunked_eval":true,"sub_50ms_latency_guarantees_for_proxy_routing":false,"custom_regex_and_deterministic_denylist_rulesets":true,"outbound_toxicity_and_competitor_mention_blocking":true,"zero_data_retention_and_prompt_privacy_guarantees":false,"hallucination_and_topical_off_brand_drift_detection":true}` | `{"iso_27001":false,"soc2_type_ii":false,"pricing_model":"free_open_source","architecture_model":"open_source_python_library","rag_retrieval_chunk_poisoning_defense":false,"prompt_injection_and_jailbreak_blocking":true,"inbound_pii_and_credit_card_masking_redaction":false,"streaming_token_interception_and_chunked_eval":false,"sub_50ms_latency_guarantees_for_proxy_routing":false,"custom_regex_and_deterministic_denylist_rulesets":false,"outbound_toxicity_and_competitor_mention_blocking":true,"zero_data_retention_and_prompt_privacy_guarantees":true,"hallucination_and_topical_off_brand_drift_detection":false}` |
| integrations.count | 7 | - |
| integrations.list | `[{"name":"OpenTelemetry"},{"name":"Snowflake"},{"name":"AWS Bedrock"},{"name":"Google Vertex AI"},{"name":"AWS"},{"name":"MCP"},{"name":"Claude Code"}]` | - |
| language.primary | - | Python |
| market.availability | `{"primaryMarkets":["US"],"availabilityScope":"global","availableCountries":[],"notAvailableCountries":[]}` | - |
| platform.support | `{"web":true}` | `{"cli":true}` |
| pricing | `{"type":"hybrid","plans":[{"free":true,"name":"Free","description":"Free version of Arthur Platform for monitoring and governing AI workloads","contactSales":false},{"free":false,"name":"Enterprise","description":"Enterprise version with enhanced performance, additional features, and capabilities","contactSales":true}],"summary":"Free tier available; enterprise pricing by quote","freeTier":true,"sourceUrl":"https://www.arthur.ai","retrievedAt":"2026-09-10T09:12:23.418Z"}` | `{"type":"open_source","summary":"Free and open source under MIT license","freeTier":true,"sourceUrl":"https://github.com/pricing","retrievedAt":"2026-08-21T10:59:49.492Z"}` |
| pricing.free_tier | yes | yes |
| pricing.model | free | open_source |
| pricing.price_level | free | free |
| pricing.starting_price | `{"amount":60,"currency":"USD"}` | - |
| pricing.transparent | yes | yes |
| security.gdpr | yes | - |
| security.scorecard | - | 5.9 |
| security.soc2 | yes | - |
| security.trust_center | - | https://github.com/security |
| security.vulnerabilities | - | `{"count":0,"source":"https://advisories.ecosyste.ms/api/v1/advisories?ecosystem=pypi&package_name=codeshield&per_page=100","last_12m":0,"max_severity":null}` |

## Capabilities (AI Guardrails)

| Capability | Arthur Shield | Llama Guard |
|---|:--:|:--:|
| **Capabilities** |  |  |
| Architecture model | Managed cloud inline proxy | Open source python library |
| Prompt injection and jailbreak blocking | ✓ | ✓ |
| Inbound pii and credit card masking redaction | ✓ | ✗ |
| Outbound toxicity and competitor mention blocking | ✓ | ✓ |
| Hallucination and topical off brand drift detection | ✓ | ✗ |
| Sub 50ms latency guarantees for proxy routing | ✗ | ✗ |
| Custom regex and deterministic denylist rulesets | ✓ | ✗ |
| Streaming token interception and chunked eval | ✓ | ✗ |
| Rag retrieval chunk poisoning defense | ✓ | ✗ |
| Zero data retention and prompt privacy guarantees | ✗ | ✓ |
| SOC2 type ii | ✓ | ✗ |
| ISO 27001 | ✗ | ✗ |
| Pricing model | Flat enterprise tier | Free open source |

*Source: vioscaleAI. Generated 2026-10-16T20:16:09.942Z. "-" = undocumented, not absent.*