KoralaRecord evidence
Facts arrive from verification sessions, from your own systems or from your own model, each with its source and age.
Fraud & risk
Catch fraud across cases. Decide with your own policy.
Korala compares each session with the ones before it, scores each customer under the risk policy you write, and keeps each decision with its inputs and reasons. For the customers you monitor, Korala reassesses on its own when evidence changes or goes stale.
The process
From evidence to a decision you can explain.
Apply your policy
A published, versioned policy turns the evidence into a score, a band and a decision: none, review, restrict or reject.
Act on the result
Actions land in one queue for your team. Signed webhooks tell your systems what changed.
Reassess
Turn on monitoring, and Korala reassesses the customer under the same policy when new evidence arrives or a review deadline passes.
Inside the product
All of a customer’s signals in one view.
The customer timeline keeps evidence, assessments and actions in order, so a reviewer sees why a case reached them.
Read the risk guide →Customer · Example
Onboarding policy v3 · Sam Chen
- 1Evidence recorded
- 2Review
- 3Decision
Same document seen before2 other sessions
NetworkData-centre address
Policy decisionReview
Fictional customer. The policy sent this case to a reviewer.
Fraud across sessions
Some fraud only shows across cases.
Reused documents
The same file, the same text or the same details under another name. Korala fingerprints every upload and compares it with earlier sessions.
Repeat identities
One person behind many accounts: the same name and date of birth, government ID, address or network turning up again.
Bursts and clusters
Too many attempts too fast, and many applicants from one IP address or subnet.
Network signals
Tor exits, data-centre and hosting networks, and an IP country that does not match the document.
Browser reuse
One browser installation used by several different customers. Opt-in. A match sends the case to a reviewer.
Device integrity
Automation tools and virtual cameras. In development. We will claim it once we have measured it.
Your policy
Write your rules and keep the reasons.
Versioned policies
Inputs, factors, bands and rules in JSON. Validate and simulate a draft, then publish it. Each assessment records the policy version that produced it.
Explanations kept
Each assessment stores the exact inputs, the score, the band and the rules that fired, so you can explain the decision later.
Overrides with a reason
An owner or admin can override a result. The override, its reason and who made it stay in the record.
Signals from your model
Send your own model's judgment as evidence. Korala accepts it only from the key you name and keeps where it came from.
An AI brief for the reviewer
When a case needs a closer look, an AI investigation brief gathers the customer's evidence, assessments and open actions, and points out saved claims about the same fact that may not fit together. Each suggestion links to its source records, and the reviewer accepts or dismisses it with a reason.
How investigations work →Review what happened
Each decision, with what led to it.
Customer timeline
Evidence, assessments, overrides and actions for one customer, in order and filterable.
Action queue
Open actions across customers in one place, with what raised each one.
Signed webhooks
Assessment and action events reach your systems signed, with delivery history to reconcile against.
Risk webhooks →With your AI assistant
Ask for it in plain words.
“Draft a policy that sends anyone on a data-centre network to review, and test it on a few scenarios.”
Publishing a policy stays with an owner. The AI brief above runs inside Korala; this is your own assistant working with Korala.
How AI assistants work with Korala →- Draft and simulate a risk policy Coming soon
- Read assessments and open actions Coming soon
Build with Korala
Use risk in your own product.
Create subjects for people, businesses, accounts or transactions, append evidence and evaluate a published policy through the API. Test keys work only on test data, so you can try a policy without touching live customers.
Common questions
Questions about fraud and risk.
Does a low risk result approve a customer?
No. A decision of none means your policy requires no restriction. It is not identity approval, and Korala does not change account permissions or block payments for you.
Does Korala score customers with AI?
No. Scores come from your published policy and the checks it reads. The optional AI brief helps a reviewer compare claims; it never changes a score or a decision.
What happens when evidence is missing or stale?
Missing evidence leaves the assessment incomplete or sends it to review, as your policy says. It never counts as a pass, and evidence older than the policy allows no longer counts.
Does using Korala make our process compliant?
Your organization defines its obligations, its policy and its review process. Korala records what you checked and what your team decided.
Try a policy in the sandbox.
Write a policy, simulate it on test scenarios and run it on test customers before anything touches live data.