Skip to main content
Find answers to the most common questions about AntifraudAI below. If you can’t find what you’re looking for, reach out to [email protected] or open a ticket from the dashboard.

​
Getting Started

Most integrations take less than a day to complete. The typical path is:
  1. Generate a sandbox API key from the dashboard.
  2. Add an API call to your checkout or event flow that sends event data and reads back the decision field.
  3. Test your integration end-to-end in the sandbox environment.
  4. Swap the base URL and API key to production.
The initial setup can be as quick as 30 minutes. Complex integrations that require custom rules, per-event-type thresholds, or webhook pipelines may take a day or two more.
No. AntifraudAI is designed to sit alongside your existing checkout or event-processing flow — not replace it. The typical pattern is:
  1. Collect the payment details as you normally would.
  2. Call the AntifraudAI /events/screen endpoint before authorizing the payment.
  3. Read the decision in the response: if allow, proceed; if review, flag for ops; if block, decline.
You do not need to change your payment processor, your front-end, or your existing fraud rules. AntifraudAI is purely an API call inserted at the right point in your server-side flow.

​
Risk Scoring

False positives can occur — especially early in your integration before AntifraudAI has built up a behavioral history for your users. Common causes include:
  • New users with no history — first-time buyers naturally have less context for the model to evaluate.
  • VPN or proxy usage — legitimate users who browse via VPN may trigger network-based signals.
  • Unusual purchase patterns — a customer’s first large-value order looks anomalous by definition.
To reduce false positives you can:
  • Whitelist trusted users by adding them to your allowlist in Settings → Rules → Allowlist.
  • Adjust your review threshold so more transactions are routed to human review instead of being blocked outright.
  • Use Cases — open a case for the transaction, mark it as legitimate, and the outcome feeds back into the model.
Yes. Every /events/screen response includes a signals array that lists the individual factors that contributed to the score, along with their weighted impact. For example:
You can also view the full signal breakdown for any historical event in the dashboard under Events → [Event ID] → Score Explanation.
AntifraudAI’s machine learning models are updated continuously. Model improvements are deployed transparently — you do not need to change your integration or update an SDK version to benefit from them. Improvements to detection accuracy, new fraud pattern recognition, and global threat intelligence updates roll out automatically to all customers.

​
Data & Privacy

AntifraudAI only processes the data you explicitly send in your API requests. There is no passive data collection — AntifraudAI does not inject JavaScript into your pages, does not use cookies, and does not track users outside of the events you send.If you want to enrich scores with device fingerprinting signals, you can optionally add the AntifraudAI Device SDK to your front-end, but this is entirely opt-in and must be explicitly configured.
All data is stored in SOC 2 Type II compliant infrastructure. By default, data is stored in the US region. Enterprise plan customers can request data residency in the EU (Frankfurt) or APAC (Singapore) regions to meet local data sovereignty requirements.
Event data is retained for 90 days by default. Extended retention periods are available on paid plans:
  • Starter — 90 days
  • Growth — 1 year
  • Enterprise — configurable; up to 7 years
You can also request early deletion of specific events or all data for a given user via the API or the dashboard.

​
Billing

Usage is billed per screened event. The sandbox environment is completely free and does not count toward your quota.For production, you can choose between:
  • Pay-as-you-go — no commitment; billed monthly at a per-event rate.
  • Monthly subscription — commit to a monthly event volume at a lower per-event rate; unused events do not roll over.
See the Pricing page for current rates.
Events continue to be processed without interruption — AntifraudAI will never stop screening events mid-month. Overage events beyond your plan’s included volume are billed at the standard per-event overage rate at the end of your billing cycle.You can set a spending cap in Settings → Billing to receive an alert (or hard-stop processing) if you approach a defined threshold.

​
Integrations

Yes. AntifraudAI works with any payment processor because it operates as an independent API call in your server-side code — it has no direct coupling to any payment gateway.The recommended pattern is to call /events/screen before you call stripe.charges.create or paypal.orders.capture. If AntifraudAI returns block, you decline the payment without ever sending it to the processor, saving on processor fees and chargeback risk.
The primary integration point is the REST API, which works with any language or framework that can make HTTP requests. Official SDKs for Python and Node.js are currently on the roadmap and will be announced via the changelog.In the meantime, the Quickstart guide includes ready-to-use code snippets for Python, Node.js, Ruby, PHP, and Go.