Adverse Media Screening (AMS)
Executive Summary
Adverse Media Screening (AMS) is a core component of Strise's compliance platform, automatically scanning and classifying news articles and media sources to identify potential risk indicators associated with companies and individuals. AMS combines data ingestion at scale, NLP-powered classification, and AI-driven summarization to deliver actionable adverse media insights.
Automation and Monitoring
AMS runs continuously as part of Strise's monitoring pipeline. When new adverse media is detected for entities in your portfolio, alerts are generated automatically -- enabling event-based compliance without manual media searches.
How It Works
AMS operates through a five-step pipeline:
1. Data Collection
Strise ingests approximately 3 million news articles daily from Opoint, a global media monitoring provider. These articles span thousands of sources across multiple languages and geographies.
Opoint is the bulk of the corpus, but not the only origin of an AMS hit. Every hit records where it came from:
| Origin | What it is |
|---|---|
| Opoint | The continuously ingested global news corpus described above. |
| Dow Jones | Adverse media from the Dow Jones data set, for teams with Dow Jones enabled. |
| AI search | An article surfaced by an AI-driven web search rather than the ingested corpus. |
| Created by user | An article a team member added manually to an entity, attributed to the user who added it. |
2. Data Normalization and NLP
Raw articles are processed through Strise's NLP pipeline, which extracts key entities (companies, persons, locations), classifies content by topic, and normalizes the data into a structured format suitable for graph integration.
3. Knowledge Graph Integration
Extracted entities and their relationships are linked to existing nodes in the Strise Knowledge Graph. This contextual linking ensures that media mentions are correctly attributed to the right companies and individuals, even when names are ambiguous.
4. Keyphrase Search
Articles are matched against entities using keyphrase-based search algorithms that identify relevant adverse media hits while filtering out noise and false positives.
5. AMS Summarization
Matched articles are summarized using Google Gemini, producing concise summaries that highlight the key risk-relevant information for reviewers.
AMS Topics
AMS classifies articles into the following topic categories:
- Core criminal acts -- Fraud, theft, embezzlement, corruption, bribery
- Violence -- Assault, murder, violent crime
- Terrorism and espionage -- Terrorism financing, espionage, state-sponsored threats
- Drug and substance -- Drug trafficking, substance-related offenses
- Corporate and economic -- Tax evasion, market manipulation, insider trading, accounting fraud
- Legal and judicial -- Lawsuits, regulatory enforcement, court proceedings
- Specialized -- Human trafficking, environmental crime, cybercrime
- Miscellaneous -- Other risk-relevant media that does not fit the above categories
New Functionality (Q3 2025)
AMS Settings Page
A dedicated AMS settings page allows teams to configure their adverse media screening parameters:
- Search Duration -- Choose how far back to search: 1 year, 2 years, 3 years, or 5 years
- Search Language -- Select which languages to include: Danish (DA), English (EN), Finnish (FI), Swedish (SV), Norwegian (NO)
- Keywords -- Define custom keywords to refine screening criteria
- Filtering by Sources -- Include or exclude specific media sources
- Filtering by Topics -- Select which AMS topic categories to screen for
Company and Person AMS
- Company and person pages, and their review pages, display up to 100 AMS hits per entity
Reviewing an AMS Hit
Clicking a hit opens the full article, alongside the context Strise has derived from it.
AI summary
Each article carries an AI summary — a short, risk-focused write-up of the article, generated on demand when you open it. Articles a team member added manually do not have one, since the summary is generated from the ingested article text.
Colour guide
Strise highlights the entity mentions it found in the article body and colour-codes them by how the entity appears in that article. The Colour guide panel beside the article lists each category with a count of how many mentions fall into it:
| Colour | Category | Meaning |
|---|---|---|
| Purple | Affected | The entity is mentioned in connection with the adverse content, but is not the subject of it. |
| Orange | Negative | The entity is the subject of the adverse content. |
| No highlight | Neutral | The mention carries no adverse signal. |
A mention classified as Negative takes precedence over Affected, so an entity counted as Negative is not also counted as Affected.
This is separate from the topic classification described under AMS Topics: topics say what the article is about, the colour guide says how this particular entity figures in it.
Marking a hit relevant or irrelevant
Each hit carries two actions, used to record a reviewer's judgement on whether the article really concerns the entity being screened:
| Action | Effect |
|---|---|
| Include as relevant hit | Confirms the hit is a true match. It stays on the entity and counts toward the entity's AMS hit count. |
| Hide as irrelevant hit | Marks the hit a false positive and hides it from the working view. |
The decision applies to the whole cluster of duplicate articles reporting the same story, not just the copy you opened, so near-identical syndicated coverage is verified once. Decisions carry across the entity page and the Review module, and update the entity's hit count in the screening matrix immediately.
Any team member with edit access to adverse media can verify a hit; it is not restricted to managers.
Contact
For questions about AMS configuration or capabilities, contact support@strise.ai.
Last updated on