Adverse Media Screener - KYC/AML Negative News Check

Screen a person or company for adverse media (negative news) for KYC/AML and enhanced due diligence. Give the actor one or more names and it returns categorized, LLM-classified hits - financial crime, fraud, corruption, sanctions, money laundering and more - each with the entity's role, a severity score, an entity-match confidence and full source provenance.

Adverse-media screening is a mandatory step in real KYC/AML and EDD workflows, and it is the one area sanctions/PEP lists do not cover. Most tools either dump raw news or bury "reputation risk" inside a broad bundle. This actor is purpose-built for adverse media and - crucially - filters out the false positives that keyword search produces.

Why the classification matters (false positives)

A naive keyword search for "Acme Corp" lawsuit flags Acme even when Acme is the one suing someone. This actor uses an LLM to read each hit and decide:

• Is the screened entity the wrongdoer? (perpetrator/defendant) - only then is it adverse. If the entity is the plaintiff, a victim, a regulator, or merely mentioned, it is not flagged.
• Is it even the same entity? Namesakes (a different person/company sharing the name) are dropped via an entity-match confidence.
• How serious is it? A severity score separates a major criminal case from a minor, old, or unproven mention.

Example: screening Patagonia Inc surfaces trademark lawsuits - but Patagonia is the plaintiff, so they are correctly not flagged as adverse. Screening Wirecard AG correctly returns fraud, money laundering and embezzlement at high severity.

Input

{
  "entityNames": ["Wirecard AG", "Jan Marsalek"],
  "entityType": "auto",
  "country": "de",
  "minSeverity": "low",
  "maxHits": 25
}

Parameter	Type	Description
entityNames	array	One or more people/companies to screen. Each is billed and returned separately.
entityType	enum	auto / person / company - improves disambiguation
aliases	array	Alternative names / transliterations (single-entity screening)
country	string	ISO code or name to bias the search and aid disambiguation
categories	array	Only return these risk categories (optional)
minSeverity	enum	low / medium / high - minimum severity to return
maxHits	integer	Max adverse hits per entity (1-50)

Output (one record per entity)

{
  "entityName": "Wirecard AG",
  "overallRisk": "high",
  "adverseHitCount": 6,
  "categoriesFound": ["fraud", "financial_crime", "money_laundering"],
  "hits": [
    {
      "title": "Wirecard Investors Drop Fraud Case ...",
      "url": "https://...",
      "source": "bloomberglaw.com",
      "publishedDate": "Nov 10, 2025",
      "snippet": "Wirecard invented fictional escrow accounts worth about $2 billion ...",
      "riskCategory": "fraud",
      "severity": "high",
      "entityRole": "perpetrator",
      "entityMatchConfidence": "high",
      "reason": "Wirecard AG is accused of inventing fictional escrow accounts."
    }
  ],
  "sources": ["Google News", "Google Web (adverse-term search)"],
  "screenedAt": "2026-06-12T...",
  "disclaimer": "Adverse-media mentions for analyst review - not verified allegations."
}

Field	Description
overallRisk	clear / low / medium / high - the worst severity found
adverseHitCount	Number of genuine adverse hits returned
categoriesFound	Distinct risk categories across the hits
hits[]	Each adverse hit: source, date, snippet, risk category, severity, entity role, match confidence, and a one-line reason
candidatesScreened	How many candidate articles were examined before filtering

Risk categories: financial_crime, fraud, corruption_bribery, sanctions, money_laundering, terrorism, organized_crime, regulatory_enforcement, litigation, environmental, other.

How it works

1. Query expansion + search - Google News plus a web search with adverse-term expansion (fraud, investigation, sanctions, corruption, money laundering, ...), via Serper. Aliases and country bias the search.
2. De-duplication - candidate hits are de-duplicated by URL.
3. LLM classification - every hit is classified for risk category, the entity's role, severity and entity-match confidence.
4. Filtering - only genuine adverse hits survive (entity is the wrongdoer, confident match, meets your severity/category filters).
5. Provenance - every hit keeps its source, URL and date for audit.

Search and the LLM are included - no API keys required.

Use Cases

• KYC / KYB onboarding - the adverse-media step alongside sanctions/PEP checks
• Enhanced due diligence (EDD) - investigate high-risk counterparties and UBOs
• Ongoing monitoring - re-screen your book periodically for new negative news
• Investigations & journalism - map allegations against a person or company

Compliance & methodology notes

• Output is "media mentions" attributed to their sources for analyst review - not verified allegations or a legal determination. Confirm against primary records before any decision.
• The actor is designed to minimise false positives (entity role + match confidence), but no automated screen is perfect - treat results as a triage aid.
• Coverage is open-web news and search; combine with sanctions/PEP and registry checks (see the suite below) for full KYC/AML.

Related Actors (regdata KYC/AML suite)

• Poland Parliamentary PEP Scraper - PEP screening dataset
• Poland CRBR Beneficial Owners - UBO/beneficial owners
• Germany Handelsregister Scraper - company data + officers
• Full suite: apify.com/regdata