What is it?
Adverse Media is Complif’s feature for identifying compliance risks (AML/CFT) for an individual or a company, based on an analysis of their presence in public media and on the web.
The process combines searches on public search engines with AI-assisted analysis, and automatically evaluates whether there are risk indicators across five key categories. The result is a structured report with the risk level per category, the sources consulted, and an executive summary. When relevant risks are detected, a case is automatically created for a compliance analyst to investigate.
Important: the tool does not issue verdicts or make automatic decisions about individuals. It identifies and escalates risks for a human analyst to review and decide.
How does it work?
The analysis runs asynchronously and in the background, without blocking onboarding or the user’s operation. It can be triggered in two ways:
Automatically, upon certain system events (for example, the creation of an onboarding or of a person).
Manually, by a user with the appropriate permissions.
The process flow, in summary:
Data validation. The system verifies that the person exists and has the minimum required data, and automatically determines the geographic scope of the search.
Search in public sources. For each risk category and each country, public web sources are queried by combining the person’s name with risk terms in the corresponding language.
AI analysis. Results are evaluated according to strict identity and risk classification rules.
Report generation. A report is produced with the analysis by category (including an executive summary of the findings), the terms searched, the countries analyzed, the cited sources, and the evidence.
Case creation, if warranted based on the risk detected.
The entire execution is logged for audit purposes.
Geographic and language scope
The search is not limited to the organization’s country. All countries relevant to the person are automatically determined:
The organization’s operating country.
The nationality (or country of incorporation, for a legal entity).
The country of birth.
The countries of addresses with approved status (non-approved addresses are ignored, to avoid operating on unverified data).
Each country is searched in its corresponding language (today: Spanish, English, or Portuguese). When the language of the country being searched doesn’t match the person’s —for example, searching in the United States but the person is from a Spanish-speaking country, or vice versa— the system combines both languages in that particular search, so as not to lose coverage.
Risk categories evaluated
Category | What it aims to detect |
Government Ties (PEP) | Public office, ties to the State, government contracts, political influence. |
Public Exposure | The person’s or company’s level of public notoriety. |
Money Laundering | Sanctions (OFAC/UN/EU), terrorist financing, offshore structures, financial scandals (Panama Papers, Pandora Papers, etc.). |
Criminal Activity | Convictions, indictments, or investigations for serious crimes. |
Economic Activity | Membership in high-risk sectors per FATF standards (casinos, real estate, crypto, jewelry, money exchange services, etc.). |
Risk scale
Level | Meaning | Action |
🔴 HIGH | Relevant indicators requiring investigation. | Generates a case (by default). |
🟠 MEDIUM | Presence in risk contexts without direct confirmation. | Configurable: may generate a case depending on the defined threshold. |
🟡 LOW | No significant indicators. | Only recorded in the report. |
How the AI evaluates risk
Before assigning any risk level, the analysis applies a mandatory identity validation process:
Identity filter. The system determines which results actually correspond to the person being analyzed, accounting for spelling variations, abbreviations, and compound surnames. A result is discarded only when it clearly belongs to a different person; when in doubt, it is included — in compliance screening, a false positive is always preferable to a false negative. The final decision rests with the human analyst.
Risk classification. Identity uncertainty never reduces the risk level: if a result was included and its content describes high-risk activity, it is classified as HIGH. For MEDIUM and HIGH levels, the classification is always backed by the specific sources that support it, so that any finding can be traced back to the concrete evidence.
Real example of a report
Value for the customer
Broad coverage without manual effort: replaces an analyst’s manual Google search by country/language/category, which is also slow and inconsistent between analysts.
Automatic multi-country and multi-language coverage: nothing needs to be configured per person; the system automatically detects which countries and languages to search based on nationality, approved addresses, and the organization’s country.
Conservative, consistent criteria (AML-first): when in doubt, it prioritizes not missing a real risk over avoiding false positives.
Full traceability: every search is logged with the exact term used, the results obtained, and the reasoning behind the analysis, which facilitates internal audits and audits by regulators/correspondent banks.
Automatic case escalation: when warranted, a compliance case is generated without manual intervention, with the risk level and evidence already attached.
High configurability: each organization can adapt the search policy to its own risk appetite.
Available configurations
Risk categories
One or more of the 5 risk categories can be excluded or disabled, either organization-wide or on a per-category basis.
Custom search terms can be defined for a category, replacing the default terms.
You can define from which risk level (low/medium/high) a specific category generates a case.
Case creation
You can define which risk levels automatically generate a case (by default, only the high level).
You can define under which case type/template the compliance case is created.
Duplicate case creation for the same person can be prevented.
Search filters
Filter results to only those published since the last successful search (useful for ongoing monitoring, avoids reprocessing what’s already been reviewed).
Filter results published within a specific date range.
Exclude specific domains from the results (for example, social media or other unreliable sites).
Name search mode
Three modes are available:
Exact: searches only the full name exactly as loaded (stricter, less noise, but may miss results with name variants).
Broad: searches the name without additional restrictions (more results, but also more noise).
All variants (default): tries reasonable combinations of the name (with/without middle name, initial, second surname, etc.) — the recommended balance between coverage and precision.