AI Model Jurisdiction Risk Scorer

Pick a model and your data class, and get a green / amber / red risk band from its jurisdiction, license, and compliance posture. For open-weight models, toggle self-hosting to see the China-jurisdiction exposure clear.

What this calculator does

Scores the data-jurisdiction risk of running a given model on a given class of data - public, internal, regulated PII, or health/financial - and returns a green / amber / red band from the model's jurisdiction, license, and published compliance posture, plus the self-host mitigation that can clear China-jurisdiction exposure.

Why use it
  • A China-hosted endpoint processes your prompts under China's National Intelligence Law and usually carries no published SOC2/HIPAA - for regulated data that is often a blocker, and a vague gut call will not survive a compliance review.
  • Open weights are mitigable: self-hosting them (or a non-China gateway region) moves the jurisdiction and can flip a red verdict to green. This shows exactly which models that works for.
For open-weight models, self-hosting (or a vetted non-China gateway region) removes the China-jurisdiction exposure.
Pick a model to score in detail

🀄 Open-weight

DeepSeek4
GLM (Z.ai)3
Kimi (Moonshot)3
MiMo (Xiaomi)2
MiniMax2
Nemotron (NVIDIA)2
Qwen (Alibaba)4
Step (StepFun)1

🛡️ Western frontier (API-only)

Claude (Anthropic)14
Cohere4
Gemini (Google)12
Grok (xAI)6
Mistral19
OpenAI37

Pick a model and a data class - the score updates instantly. Dots show each model's band at the current sensitivity.

Risk verdictNot recommended - Z.ai: GLM 5.2 · Regulated PII (GDPR/CCPA)
Z.ai: GLM 5.2 GLM (Z.ai)
Regulated PII (GDPR/CCPA)
Not recommended
Data class: Regulated PII (GDPR/CCPA). Personal data under privacy law; cross-border processing is a compliance event.
Provider endpoint is processed under China jurisdiction (National Intelligence Law applies).
No published SOC2/HIPAA, and the data class is regulated (PII/PHI).
Recommendation: Not recommended as a direct China-hosted endpoint for this data class. Self-host the open weights (or a vetted non-China gateway) to bring it to amber, or choose a model with a published compliance posture.
Jurisdiction*China (National Intelligence Law applies to provider-hosted endpoints)
License*MIT
Compliance*no published SOC2/HIPAA; China-hosted by default; content restrictions on sensitive topics
Data residency*China (direct) / varies via gateway
Independent verification*partial
This data class carries a compliance overhead. Estimate the annual dollar cost in the AI compliance cost calculator →
All models at this sensitivity 19 red 0 amber 94 green
ModelCompanyTypeRisk
Deepseek Chat DeepSeek open-weight red
DeepSeek V3.2 (reasoner) DeepSeek open-weight red
DeepSeek V4 Flash DeepSeek open-weight red
DeepSeek V4 Pro DeepSeek open-weight red
Z.ai: GLM 4.7 Flash GLM (Z.ai) open-weight red
Z.ai: GLM 5 GLM (Z.ai) open-weight red
Z.ai: GLM 5.2 GLM (Z.ai) open-weight red
Kimi K3 Kimi (Moonshot) open-weight red
MoonshotAI: Kimi K2.6 Kimi (Moonshot) open-weight red
MoonshotAI: Kimi K2.7 Code Kimi (Moonshot) open-weight red
Xiaomi: MiMo-V2.5 MiMo (Xiaomi) open-weight red
Xiaomi: MiMo-V2.5-Pro MiMo (Xiaomi) open-weight red
MiniMax: MiniMax M2.7 MiniMax open-weight red
MiniMax: MiniMax M3 MiniMax open-weight red
Qwen: Qwen3 Coder Plus Qwen (Alibaba) open-weight red
Qwen: Qwen3.6 Flash Qwen (Alibaba) open-weight red
Qwen: Qwen3.7 Max Qwen (Alibaba) open-weight red
Qwen: Qwen3.7 Plus Qwen (Alibaba) open-weight red
StepFun: Step 3.7 Flash Step (StepFun) open-weight red
Claude Haiku 4.5 Claude (Anthropic) API-only green
Claude Opus 4.1 Claude (Anthropic) API-only green
Claude Opus 4.5 Claude (Anthropic) API-only green
Claude Opus 4.6 Claude (Anthropic) API-only green
Claude Opus 4.7 Claude (Anthropic) API-only green
Claude Opus 4.8 Claude (Anthropic) API-only green
Claude Sonnet 4.5 Claude (Anthropic) API-only green
Claude Sonnet 4.6 Claude (Anthropic) API-only green
Claude Sonnet 5 Claude (Anthropic) API-only green
claude-fable-5 Claude (Anthropic) API-only green
claude-haiku-3-5 Claude (Anthropic) API-only green
claude-mythos-5 Claude (Anthropic) API-only green
claude-opus-4 Claude (Anthropic) API-only green
claude-sonnet-4 Claude (Anthropic) API-only green
Command Cohere API-only green
Command Light Cohere API-only green
Command R Cohere API-only green
Command R+ Cohere API-only green
Gemini 2.5 Flash Gemini (Google) API-only green
Gemini 2.5 Flash-Lite Gemini (Google) API-only green
Gemini 2.5 Pro Gemini (Google) API-only green
Gemini 3 Flash Gemini (Google) API-only green
Gemini 3.1 Flash-Lite Gemini (Google) API-only green
Gemini 3.1 Pro Gemini (Google) API-only green
Gemini 3.5 Flash Gemini (Google) API-only green
gemini-2.5-flash Gemini (Google) API-only green
gemini-2.5-flash-lite Gemini (Google) API-only green
gemini-2.5-pro Gemini (Google) API-only green
gemini-3.1-flash-lite Gemini (Google) API-only green
gemini-3.5-flash Gemini (Google) API-only green
Grok 4 (legacy) Grok (xAI) API-only green
Grok 4.1 Fast (non-reasoning) Grok (xAI) API-only green
Grok 4.1 Fast (reasoning) Grok (xAI) API-only green
Grok 4.3 Grok (xAI) API-only green
grok-4-5 Grok (xAI) API-only green
grok-build-0-1 Grok (xAI) API-only green
codestral Mistral API-only green
devstral-2 Mistral API-only green
devstral-small-2 Mistral API-only green
magistral-medium Mistral API-only green
magistral-small Mistral API-only green
ministral-3-14b Mistral API-only green
ministral-3-3b Mistral API-only green
ministral-3-8b Mistral API-only green
Mistral Large 3 Mistral API-only green
mistral-medium-3-5 Mistral API-only green
mistral-nemo Mistral API-only green
mistral-small-4 Mistral API-only green
mistral-small-creative Mistral API-only green
mixtral-8x22b Mistral API-only green
mixtral-8x7b Mistral API-only green
voxstral-mini Mistral API-only green
voxstral-small Mistral API-only green
voxtral-mini Mistral API-only green
voxtral-small Mistral API-only green
NVIDIA: Nemotron 3 Super Nemotron (NVIDIA) open-weight green
NVIDIA: Nemotron 3 Ultra Nemotron (NVIDIA) open-weight green
GPT 5 Chat OpenAI API-only green
GPT 5 Codex OpenAI API-only green
GPT-5 OpenAI API-only green
GPT-5 Mini OpenAI API-only green
GPT-5 Nano OpenAI API-only green
GPT-5 Pro OpenAI API-only green
gpt-5-3-codex OpenAI API-only green
gpt-5.1 OpenAI API-only green
GPT-5.1 OpenAI API-only green
gpt-5.1-codex OpenAI API-only green
gpt-5.1-codex-max OpenAI API-only green
gpt-5.1-codex-mini OpenAI API-only green
gpt-5.2 OpenAI API-only green
GPT-5.2 OpenAI API-only green
GPT-5.2 Pro OpenAI API-only green
gpt-5.2-codex OpenAI API-only green
gpt-5.2-pro OpenAI API-only green
gpt-5.3-codex OpenAI API-only green
gpt-5.4 OpenAI API-only green
GPT-5.4 OpenAI API-only green
GPT-5.4 Mini OpenAI API-only green
GPT-5.4 Nano OpenAI API-only green
GPT-5.4 Pro OpenAI API-only green
gpt-5.4-mini OpenAI API-only green
gpt-5.4-nano OpenAI API-only green
gpt-5.4-pro OpenAI API-only green
gpt-5.5 OpenAI API-only green
GPT-5.5 OpenAI API-only green
GPT-5.5 Pro OpenAI API-only green
gpt-5.5-pro OpenAI API-only green
gpt-5.6 OpenAI API-only green
gpt-5.6-luna OpenAI API-only green
gpt-5.6-sol OpenAI API-only green
gpt-5.6-terra OpenAI API-only green
o1 Pro OpenAI API-only green
o3 Mini OpenAI API-only green
o3 Pro OpenAI API-only green

Risk band is a heuristic from data sensitivity, model jurisdiction, and published compliance posture, not legal advice. All governance fields (jurisdiction, license, compliance, residency, verification) are curated/inferred by aicost.ai* and reflect the default provider endpoint, not vendor-official guarantees. Self-hosting open weights or routing via a non-China gateway region changes the jurisdiction and the band. Any prompt sent to a China-hosted provider endpoint is processed under China's National Intelligence Law.

Go deeper

Our playbooks on cutting this number.

📋
AI Compliance Cost
Annual governance overhead in dollars
🌏
Chinese vs Western Cost
The price spread with the risk caveat
🔍
Open-Source Model Finder
Filter open-weight models by license + more
🖥️
Self-Host Inference
Cost of self-hosting to clear jurisdiction

Need help using this calculator for your workloads?

AICost.ai has 50+ calculators and playbooks. Schedule an AvatarVA meeting and we'll work through your real cost scenarios across AI & Cloud: visibility, cost reduction, optimization, forecasting and capacity planning, without sacrificing accuracy or performance.

📅 Schedule an AvatarVA meeting →