Product
One API to screen people against every sanctions list
AutoCompliance combines official sanctions data, an AI matching engine built for names, and an API your developers will enjoy integrating.
Single screening
Synchronous POST /v1/screenings with a p95 latency target under 350 ms. Built for onboarding and payment flows.
Batch screening
Up to 100,000 records per batch via JSON, CSV or XLSX. Paginated results, a "matches only" filter and CSV/Excel/JSONL export.
Explainable results
Every match shows the retrieval score, the verifier score, the date-of-birth signal, the alias that matched and the full list entry.
List versioning
All official lists are refreshed daily. Every version is kept, and each result references the exact entry version.
Developer-first API
REST with an OpenAPI 3.1 spec, idempotency keys, cursor pagination, RFC 9457 errors and HMAC-signed webhooks.
Retention control
Set retention per organisation or switch on zero-retention mode for stateless screening.
Matching engine
Built for how names actually vary
Name screening has two failure modes. Missing a sanctioned person is a regulatory breach. Flagging too many innocent customers costs analyst time and conversion. AutoCompliance addresses both with two stages.
Stage 1: fuzzy phonetic retrieval
Every name in every list is encoded as a vector of character n-grams and indexed for approximate nearest-neighbour search. Spelling variants, transliterations, typos and reordered name parts land close to each other, so candidates are found in milliseconds without generating thousands of rule-based variations.
Stage 2: AI verification
A language model compares the input with each candidate and estimates whether they are the same name. It recognises diminutives and national forms, and rejects strings that look similar but are different names.
Scoring
The final 0–100 score combines both stages with a fuzzy date-of-birth comparison that understands year-only and range dates. Single-word names receive a penalty unless the date of birth confirms them. You set one threshold per project.
What the engine handles
- Transliteration and phonetic spelling (Mohammed / Muhamad)
- Name order (Ivanov Ivan / Ivan Ivanov)
- Missing parts (no middle name or patronymic)
- Diminutives and national forms (Alexander / Sasha)
- Typos, extra spaces and punctuation
- Titles and honorifics (Dr, Mr, Sir) ignored
- Double surnames matched on either part
What it keeps apart
- ✕Maria ≠ Marina
- ✕Alexander ≠ Alexey
- ✕Inna ≠ Anna
Input names are currently accepted in Latin script. Matching quality is measured on a benchmark of perturbed list entries and control names before every model release.
API
Integrate in an afternoon
Test keys work against sandbox lists with documented hit and no-hit names, so you can write automated tests before going live.
- Bearer API keys with separate test and live modes
- Idempotency keys on every write
- HMAC-signed webhooks with automatic retries
- Per-key rate limits with standard headers
- Problem Details (RFC 9457) errors with stable codes
curl https://api.auto-compliance.com/v1/batches \
-H "Authorization: Bearer ak_test_..." \
-H "Idempotency-Key: 6f1c2e0a-batch-2026-11-20" \
-F [email protected]
# → 202 Accepted { "id": "bat_01JD...", "status": "queued", "total_items": 48210 }Roadmap
Coming next
Merchant portal
Web screening, batch uploads, alert review with audit trail, reports and PDF screening certificates.
Custom lists & whitelist
Your own blocklists and suppression of confirmed false positives.
Bring your own provider
Index licensed vendor data next to official lists at a reduced price.
Ongoing monitoring
Automatic re-screening of your customer base when lists change.
PEP & criminal lists
Additional risk categories beyond sanctions.
Legal entities
Company name screening with legal-form normalisation.
Screen smarter from day one
We are onboarding a small group of design partners. Tell us about your screening volume and lists, and we will set up sandbox access.