FRAUD & SCAM DETECTION
Generally availableStop the payment diversion before the wire goes out.
200 fraud-scheme labels across 39 languages - payment diversion, impersonation, fake invoices, social engineering - each returned with the configuration ID and receipt you need to reproduce the call. No generative model is used to produce classifications.
200 labels · 4 fraud groups · 39 languages · p95 target <500ms · No generative model is used to produce classifications
Re-run with the same pinned version → identical result
What this analyzer does and doesn't do
- Detect 200 labeled fraud and deceptive-scheme behaviors in text
- Return confidence-ranked signals your fraud engine can route on
- Cover 39 languages
- ✕Verify bank details, IBANs, or transaction records
- ✕See account history, invoices on file, or customer metadata
- ✕Determine whether activity is actually criminal
- ✕Replace your fraud analysts or make the final call
Sentiqra returns structured signals. Your policy, reviewers, and models make the decision. See the full boundary statement →
WHAT YOU'LL DETECT
Not "fraud: 0.91". The specific scheme - with the label your playbook already knows.
Payment Diversion: BEC & CEO Fraud
pd_bec_invoice_redirect
“Please update the beneficiary account for this invoice” → very high
Impersonation: Authority Fraud
imp_gov_official
“This is the tax authority - you have an outstanding liability” → very high
Social Engineering: Tech Support Scams
se_remote_access_tech
“Install this app so our technician can fix the issue” → very high
Fake Invoices: Document Forgery
fi_bank_detail_altered
Detected - example withheld
Investment & Crypto Pump Scams
se_pig_butcher_crypto
Detected - example withheld
200 indicators across 4 fraud groups - covering 39 languages.
Same deterministic product contract as every Sentiqra analyzer: pinned configuration, signed receipt, replayable results. Per-language benchmark publication is pending for this analyzer - evaluate it on your own labeled workload; the published methodology applies when results ship.
WHERE IT GETS USED
Three jobs this analyzer does well.
Vendor payment-diversion screening
An AP team screens every "updated bank details" email through Fraud & Scam before touching the vendor master file. pd_vendor_bank_change and fi_modified_invoice hits put the payment on hold pending out-of-band verification - the classic BEC kill-chain stopped at step one.
Fit signals: payment-instruction changes · supplier email volume · out-of-band verification policy
Executive and authority impersonation triage
Inbound mail claiming a CEO, tax authority, or law-enforcement request runs through the impersonation families. pd_bec_ceo_request combined with se_urgency_pressure routes straight to the fraud desk instead of an employee's judgment call.
Fit signals: BEC exposure · wire authority · urgency + secrecy patterns
Member scam warnings on platforms
A marketplace runs buyer-seller messages through the social-engineering families. When se_pig_butcher_crypto or se_refund_scam fires, the member sees an in-context warning before money or credentials move - protection without a silent ban.
Fit signals: P2P messaging · member financial losses · warning-friction UX
Send text. Get ranked labels + a receipt you can replay.
POST /v1/analyze
Authorization: Bearer sq_live_your_api_key_here
{
"text": "Urgent - our bank account has changed. Please re-route this month's invoice payment to the new IBAN attached.",
"language": "en",
"analyzers": ["fraud-scam"]
}{
"language": "en",
"signals": [
{ "analyzer": "fraud-scam", "code": "payment_diversion", "label": "Payment Diversion", "score": 0.9, "confidence": "very_high", "level": "category" },
{ "analyzer": "fraud-scam", "code": "pd_bec_ceo_request", "label": "CEO Urgent Payment Request BEC", "score": 0.9, "confidence": "very_high", "level": "label", "parent": "payment_diversion" },
{ "analyzer": "fraud-scam", "code": "pd_bec_attorney_confidential", "label": "Attorney Confidential Payment BEC", "score": 0.9, "confidence": "very_high", "level": "label", "parent": "payment_diversion" }
],
"meta": { "analyzers": [{ "name": "fraud-scam", "status": "completed", "signal_count": 3, "billed": true }], "partial": false },
"usage": { "analyzer_runs": 1, "billed": true },
"verification": {
"configuration": "sqcfg_CQ8CN62V",
"fingerprint": "fp_8a72...19f",
"input_commitment": "ic_31bd...a42",
"signature": "sig_ed25519_4c1f...e2a"
}
}READING THE OUTPUT
What a Fraud & Scam Detection result tells you - and what it doesn't.
Intended task
Flag documented fraud and scam patterns - payment diversion, impersonation, advance-fee lures - in text before money or credentials move.
Input requirements
text- the only field analyzed; up to 4,096 Unicode code points. Over the limit is rejected withinput_too_long, never truncated.language- must be a language this analyzer supports. Unsupported codes are rejected with400 unsupported_language, never analyzed as a different language.analyzers- include"fraud-scam"; up to 2 analyzers per request, one billed run each.- No context beyond
text- no user IDs, history, profiles, or attachments are read or inferred. - Analyzes the submitted text only - sender identity, headers, and account relationships are not inputs.
Ordinary positive
“Please update our payment details - our account is under audit, so wire this month's invoices to the new account below.”
Safe inference: The text matches documented payment-diversion patterns - hold the payment change and verify out-of-band before any wire.
Confusing negative · Quotation
“Security bulletin: watch for emails like "wire this month's invoices to the new account below" - report them to finance.”
What happens: Payment-diversion signals can fire on the quoted lure. Training material, bulletins, and forwarded suspect emails all contain the same language as the attack itself - the signal describes the text, not whether this message is itself an attack.
Why it's tricky: A bare request like "can you send me $200?" may also be too thin to match strongly - fraud patterns are contextual. Treat flagged text as a verify-first trigger, and unflagged payment changes as "not detected" rather than verified-safe.
What you can safely infer
- Which documented fraud patterns the text matched, and at what signal-match strength.
- That the same message on the same pinned configuration replays identically - the receipt supports your audit trail.
What you must not infer
- ✕That a flagged message proves fraud or wrongdoing - signals route verification, they do not determine guilt.
- ✕That the sender is a criminal - impersonation signals describe the language, not the person.
- ✕That an unflagged payment request is verified-safe - always confirm payment changes out-of-band regardless of the result.
Boundary: Fraud signals are not proof of wrongdoing. They flag language consistent with documented schemes so an analyst or control step can verify - payment changes should always be confirmed out-of-band regardless of the result.
Fraud & Scam Detection is on every plan. Even Free.
Free
5,000 runs/mo
Build
$149/mo · 50,000 runs/mo
Growth
$349/mo · 150,000 runs/mo
Production
$799/mo · 500,000 runs/mo
Scale
$2,999/mo · 2,500,000 runs/mo
Enterprise
By agreement · SOC 2 Type II in progress
Deterministic by construction, provable by receipt.
A signed verification receipt - configuration ID (sqcfg_...), result fingerprint (fp2_...), input commitment (ic3_...), and Ed25519 signature (sig_ed25519_...) - ships with every successful analysis result. Pin it to the stored decision and a later replay under the same supported configuration and output profile yields the identical canonical result - the difference between "the model said so" and evidence.
Where it fits in your pipeline
Often paired with: Dating Safety for romance-fraud conversations, Cybersecurity for phishing and credential theft, Trust & Safety for general platform abuse.
Used in: Phishing & Account Protection and Marketplace & Messaging Safety
Try it on your hardest fraud cases.
5,000 analyzer runs/month · No card required
Analyzer FAQ
New to Sentiqra? It's a deterministic text-analysis API - you send text, you get labeled, confidence-scored signals plus a replayable receipt. Fraud & Scam Detection is the fraudulent scheme detection analyzer.
What does Fraud & Scam Detection detect?
Detect payment diversion, impersonation, advance-fee fraud, social-engineering lures, and scam patterns in multilingual text.
Which languages does it cover?
39 supported languages. Per-language benchmark publication is pending for this analyzer - evaluate it on your own labeled workload before relying on a specific language. Coverage is per-analyzer, not platform-wide.
How do I call it?
POST /v1/analyze with "analyzers": ["fraud-scam"]. Same endpoint, auth, and response shape as every other analyzer - you can pair it with one more analyzer in the same request.
What does a call cost?
One analyzer run per call, billed against your plan's included runs (overage per plan). All GA analyzers are on every plan, including Free - see pricing.
How do I evaluate it on my data?
Label a few hundred of your own examples, run them through the free tier, and compare our signals to your labels. The evaluation playbook walks through building that test set - it's the same method we use internally.