AUTHENTICATION
Use your existing AI7 credential
Send your personal AI7 API key as a Bearer token. The media-source credential, query strategy, and private configuration remain server-side.
Authorization: Bearer YOUR_AI7_API_KEYREQUEST
Screen a person or business
entity_typeenum · required
individual or business.
namestring · required
Primary individual or business name.
aliasesstring[] · optional
Alternate spellings, former names, or transliterations.
countriesstring[] · optional
Up to 20 ISO two-letter country codes used for identity context and focused regional coverage.
languagesstring[] · optional
Up to five languages. Defaults to English and Arabic. Chinese and Russian also activate their matching country pack.
date_from / date_todate · optional
Inclusive publication-date search window.
high_risk_threshold_percentnumber · optional
Scores at or above this percentage are High Risk. Default: 90.
review_threshold_percentnumber · optional
Scores from this percentage up to the High Risk boundary require Review. Lower scores are Safe. Default: 65.
search_plusboolean · optional
Actively submits site-restricted searches for selected high-value publishers, prioritising requested countries, and consolidates the results with standard evidence.
max_resultsinteger · optional
One to 50 results. Default: 25.
deep_researchboolean · optional
Enables asynchronous screening and the optional callback workflow.
callback_urlHTTPS URI · optional
Available only with deep research. The signed final result is posted to this public URL. Partial results remain explicitly inconclusive and require review.
customer_logo_urlHTTPS URI · optional
PNG or JPEG logo shown on the automatically generated audit PDF. The PDF is permanently retained and returned through an unguessable public capability URL.
RESPONSE
Customer-ready evidence
Every retained result includes its original URL, evidence excerpt, risk category, legal stage, and combined risk score. Internal search, model, cost, and scoring diagnostics are never returned.
FieldTypeDescription
screening_idUUIDUnique screening reference.summary.outcomestringPotential match, no material results, or inconclusive when required processing is incomplete.summary.risk_levelstringHigh, review, safe, no material results, or indeterminate for an incomplete run.summary.thresholdsobjectThe applied High Risk and Review percentages and the Safe boundary.summary.risk_distributionobjectCounts of retained results in the High Risk, Review, and Safe bands.summary.categoriesobjectCount of results by adverse-event category.statusstringcompleted only when every configured search and AI review stage succeeds; otherwise partial.summary.threshold_statusstringHigh risk, review required, safe, or incomplete coverage.summary.decision_eligiblebooleanFalse whenever the run is partial and must not be treated as a completed compliance decision.coverageobjectOverall completeness, geographic scope, languages, dates, search time, and explicit limitations.coverage.sourcesobject[]Provider-neutral status for current news, open web, Search+ publisher search, archive, and official-record coverage.coverage.reason_codesstring[]Safe diagnostic codes explaining incomplete coverage without exposing suppliers.coverage.retryablebooleanTrue when retrying may complete the screening.coverage.search_plusobjectWhether Search+ ran, batch counts, catalog size, selected coverage strategy, and approved-publisher matches.coverage.search_plus.publishers_screenedobject[]Exact publisher name, domain, country, and category included in successfully completed site-restricted index queries.coverage.search_plus.publishers_not_completedobject[]Publishers assigned to a batch that failed after controlled retries.coverage.search_plus.matched_publishersobject[]Approved publishers represented in the retained customer-facing results.results[].titlestringArticle or page title.results[].urlURIOriginal evidence URL.results[].source_typestringNews index, open web, Search+ publisher search, global archive, or official regulatory evidence.results[].excerptstringRelevant indexed excerpt.results[].identity_matchobjectDeterministic identity score, match type, matched input, and customer-readable explanation.results[].search_plus_matchbooleanTrue when the evidence URL belongs to the approved Search+ publisher catalog.results[].categoriesstring[]Detected risk categories, including financial crime, enforcement, sanctions, cybersecurity, corporate misconduct, founder/leadership disputes, workforce restructuring, and financial distress. A restructuring mention is an adverse event lead, not an allegation of fraud.results[].legal_stagestringMedia mention, allegation, investigation, charge, arrest, conviction, or regulatory action.results[].risk_scorenumberCombined, explainable prioritisation score.results[].review_requiredbooleanTrue when the retained result meets the configured review threshold.reportobjectPermanent report ID, transaction ID, SHA-256 digest, and public capability URL. Anyone with the unguessable link can download the PDF.Status codes
200Screening returned; inspect
status and summary.decision_eligible before use202Deep research durably queued for signed callback delivery
400Invalid request fields
401Missing or invalid AI7 API key
502Media indexes temporarily unavailable
503Permanent report generation or storage temporarily unavailable
METHODOLOGY
Designed for review, not automated guilt decisions
The engine normalizes names and aliases, generates tightly constrained name transliterations for selected non-English languages, and performs a bounded identity-resolution pass for multi-part individual names to discover explicit aliases and associated organizations. It searches current and archived indexes with controlled retries, removes duplicate URLs, classifies adverse-event terms, and ranks evidence. The taxonomy distinguishes corporate misconduct from founder/leadership disputes, workforce restructuring, and financial distress; a layoff or founder split is a reviewable event, not automatically a fraud allegation. Search+ reserves part of the fixed 12-query guardrail for explicit publisher-domain batches, prioritises requested countries and high-value editorial categories, and reports the successful and incomplete publisher lists. A successful site-restricted index query proves the query ran; it does not prove the index contained every article published by that outlet. Partial individual-name matches are rejected before classification, while only high-confidence evidence mismatches can suppress a candidate. Any exhausted source or required classification failure makes the run partial, returns an inconclusive outcome, blocks decision eligibility, and requires manual review. A media result is a review lead—not proof of wrongdoing.
LIVE TESTER
Run an adverse media check
The cost guardrail limits each uncached screening to 12 paid search requests plus one bounded AI classification batch. Repeating the same request within 15 minutes uses the cached evidence.
API RESPONSEReady
{
"message": "Your JSON response will appear here."
}