Connect your AI model, and ARIA Shape B runs a curated battery of behavioral tests against it — on demand and on a schedule you choose — then hands you a score and a downloadable report. No integration project. No consultant. Just answers about how your model behaves.
ARIA Shape B is the standalone entry point to the ARIA platform. It takes the same behavioral evaluation engine that powers ARIA's enterprise compliance product and makes it available on its own — so any team running an AI model can start measuring its behavior today, without adopting the full platform first.
Point Shape B at your own AI/LLM endpoint. Your credentials are encrypted at rest and used only to run your tests.
Trigger a run on demand, or set a recurring schedule so your model is re-tested automatically as it changes.
Receive a behavioral score and a downloadable report you can share with your team, auditors, or customers.
Shape B is one way into ARIA — the same engine, offered standalone, that grows with you into full continuous compliance.
ARIA meets your AI wherever it lives, through three integration shapes. Shape B is the one most teams start with.
Your own pipeline posts its AI-testing results into ARIA — for teams who already run safety tests in CI.
ARIA tests your live model endpoint directly, on demand and on a schedule — behavioral evaluation with a score and report.
Real-time enforcement in the request path — layered gates screen every call before and after your model responds.
Start with Shape B, grow into the rest. When you're ready for continuous compliance, real-time guardrails, or multi-framework reporting, your Shape B account upgrades into the full ARIA platform — no migration, nothing to rebuild.
Pricing is simple: $30 per batch, per month. One batch covers up to 4 AI models and 20 evaluation runs a day. Need more? Add batches — anytime. Prepay a longer term and save.
What's a batch? A batch = up to 4 connected models + 20 runs/day (one scheduled run per model per day). 1–4 models = 1 batch. 5–8 = 2 batches. In general, batches = your models ÷ 4, rounded up. One account (and its whole team) can hold as many batches as you need.
By subscribing you agree to the Terms, Refund & Privacy policy.
To run tests on the schedule you set, we store the credential you provide for your model endpoint — encrypted at rest, and decrypted only in memory at the moment a test runs. We do not retain the raw test prompts or your model's raw responses beyond each run; we keep the derived score and timestamps so you can see how your model's behavior trends over time.
Shape B is a testing tool, not a certification. A score tells you how your model behaved on the tests we ran — it doesn't certify that your AI is compliant, safe, or free of issues. You remain responsible for your systems and for decisions you make based on the results.
Connect a model and get your first behavioral report. Questions first? We're happy to talk.
Get started Talk to us