When UBS Financial Services onboarded the companies of a customer that FinCEN's order calls Individual C, its negative news screening returned thousands of hits across more than 300 articles. A later periodic review returned more than 150 articles, and staff read only the first 25. One story said the fortune came from Russia's "loans for shares" auctions; the file dismissed it as "only speculations" because no legal proceeding had been filed. The accounts, rated higher risk from the start, held more than $175 million by February 2022. On 3 August 2026 FinCEN assessed a $125 million penalty, and the consent order says the firm "did not appropriately explain or justify" how it dispositioned the news it read.
The screening found the stories. The failure came at the review stage of adverse media screening: too many articles to read, and no recorded reason for setting them aside. That is the test this page applies to ten tools.
What do regulators expect from adverse media screening?
Table 1. What the rule texts say about adverse media (November 2026)
| Jurisdiction | Text | What it says | What the tool must show |
|---|---|---|---|
| EU | EBA/GL/2021/02, customer risk factors | "Are there adverse media reports or other relevant sources of information about the customer ... If so, are these reliable and credible?" | the source of each hit and how credible it is |
| EU | EBA/GL/2021/02, Guideline 4.64 (enhanced due diligence) | EDD may include assessing "any negative allegations" against the customer, with "adverse media searches" named as an example | deeper searches on the names that reach EDD |
| US | FinCEN Consent Order No. 2026-02 (UBS) | the independent review must test that customer risk profiles take in "information identified by UBSFS in media reports" about alleged ties to illicit activity | hits that land on the customer's risk profile, not in a separate inbox |
| UK | Money Laundering Regulations 2017, reg 33(5) | EDD may include "seeking additional independent, reliable sources" and measures to understand "the background, ownership and financial situation of the customer"; media are not named | independent sources searched and recorded on the high risk file |
| EU (from AMLR) | AMLA final draft standards, 1 October 2026 | customer due diligence standards sent to the Commission, to apply six months after entry into force; the release names PEP screening, not adverse media | a setup that can change when the final text is published |
Source: the texts linked in the table, opened 11 November 2026.
The EBA line that does the most work sits under the reputation factors: firms judge an allegation by "the quality and independence of the source" and "the persistence of reporting", and "the absence of criminal convictions alone may not be sufficient to dismiss allegations". Read against the UBS file, that is the exact reasoning the order faulted.
For a buyer, two things follow. Source tier and allegation weight are policy decisions, so a tool should let the firm set them and show the setting in the record. And every dismissal needs a reason the next reviewer can read.
How do you choose adverse media screening software?
Nine checks decide most shortlists. Each comes with the reason it matters.
- Sources, by type. Count is only half of it: ask how many sources are news, regulatory notices, court records, social platforms and forums, because early signals rarely break in a national newspaper first.
- Languages. A local fraud is reported first in the local press, in the local language. Ask how many languages are read natively and whether the original text is kept beside the translation.
- Refresh. Ask how often new articles enter the data and whether each screening also runs a live query. A monthly batch misses the story that broke last week.
- Matching and namesakes. News carries a name and a city, rarely a date of birth. Ask which identifiers the tool uses to separate a customer from a namesake, since that decides the false positive and false negative rate more than any model.
- Categories by offense. A hit should be filed by the crime it describes, such as fraud, bribery or organized crime, not by the tone of the article. Ask to see the category list and whether it can be narrowed for your sector.
- Allegation weight and lookback. An arrest, a charge and a conviction carry different weight, and a story from 15 years ago is not last month's. These two settings are where false positive control actually lives.
- Curated profiles or raw news. Research teams turn articles into reviewed profiles; machine pipelines screen raw news against your filters. One is narrower and slower, the other wider and noisier.
- Evidence and dismissals. The UBS order turned on dismissals with no recorded reason. Ask what the file keeps for each hit: source, date, language, the passage that matched, the reviewer's reason.
- Delivery and monitoring. Check the API, the batch option and how ongoing monitoring in AML starts and stops per customer, because a book of customers needs screening that keeps running after onboarding.
The 10 best adverse media screening tools compared
complead publishes this comparison and lists itself first. The other nine follow in A to Z order. They are the names the AI answers and the top search results for this question repeat most often across eleven published lists, kept only where the vendor's own adverse media page could be read on 11 November 2026. Every cell below quotes or paraphrases that page and carries the read date in the caption; "not stated" means the page read does not say, not that the product lacks it. No prices and no review scores appear.
Table 2. 10 adverse media screening tools compared (vendor pages read 11 November 2026)
| Vendor | Media sources (as stated) | Languages (as stated) | Refresh (as stated) | Categorization (as stated) | Curated profiles or raw news | Delivery |
|---|---|---|---|---|---|---|
| complead | 300,000+ articles a day, 190+ countries; social platforms and forums scored as their own tier | 30+; up to three per screening; original text kept alongside translation | 300,000+ articles processed a day | by the crime described (fraud, bribery, financial crime, organized crime), not tone | live news screened against the firm's own settings | API, platform; monitoring started and stopped per name |
| Castellum.AI | 200,000+ media sources, 220+ countries and territories | 20 | feed updates every 5 seconds; 20+ years of history | 100+ categories | own enriched data with category filters | API |
| ComplyAdvantage | over 11,000 verified national, regional and local media sources, 200+ countries and territories | native language processing; count not stated | updated every day | not stated on the page read | expert curated profiles with article snippets | API; always on monitoring |
| Dow Jones Risk & Compliance | Factiva: 33,000+ sources, 200 countries | 33 (Factiva) | not stated | regulatory, financial crime, reputational and labor risk categories | structured adverse media entity and Special Interest Persons data, plus Factiva news | feeds, APIs, partner integrations |
| LexisNexis Risk Solutions | thousands of news media sources | 59 | not stated for news | high probability matches categorized for priority | 8M+ curated WorldCompliance profiles plus news | Bridger: public or private cloud, on premise |
| LSEG World-Check | not stated | not stated | not stated | categories, subcategories, Special Interest Categories | structured media in World-Check records; Media Check for wider news (add on) | feeds, APIs, World-Check One |
| Moody's | 14,000+ trusted media sources; 3+ billion articles screened | foreign language media in 100+ countries, English summaries | not stated | 80+ risk codes | sources reviewed by researchers | data in screening workflows |
| Owlin | over 3 million sources | multiple; count not stated | continuous, "No batch updates" | risk signals classified in context, each linked to its article | raw news on any business with an online presence | API and integrations |
| Ripjar | 6 billion+ articles; about 6 million new a day | NLP in 22+; entity recognition in 50 | profiles updated twice a day | a risk summary per entity profile | machine built profiles, up to 14 million entities | not stated on the page read |
| Sigma360 | 730,000+ sources | 120+ | real time coverage | by risk type and event topic; materiality scoring | related articles grouped into one story | API |
Sources: the vendor pages linked in each entry below, read by curl on 11 November 2026; complead from its fact sheet rows, which match its live adverse media data and product pages.
Adverse media screening vendors, one by one
complead
complead reads 30+ languages across 190+ countries and processes 300,000+ articles a day, as its adverse media data page states. Each hit is assigned to a risk category by the crime described, for example fraud, bribery, financial crime or organized crime, rather than the article's tone. One screening can search a name in up to three languages at once, and the original text is kept alongside the translation. In complead's adverse media screening, offense categories, minimum source tier, lookback period and allegation weighting are "settings on your side", and social platforms and forums are scored as their own tier.
The product page's case block reads: "A payments firm caught a corruption story its old tool never saw", with "2 Relevant of 40 hits" and "0 Namesakes in the result". Ongoing monitoring starts and stops per name through API triggers. Data the firm already pays for stays in use: a licensed provider such as Moody's or Dow Jones connects with the firm's own key, and its hits share one queue with complead's. Around each customer, the connections graph follows up to six hops by default toward sanctioned, PEP or adverse media entities, each hop sourced. Ask to set one lookback and one allegation weight, then see what leaves the queue.
Castellum.AI
Castellum.AI screens adverse media on its own enriched data: its adverse media page lists 200,000+ global media sources across 220+ countries and territories in 20 languages. The feed updates every 5 seconds, 250,000 records are added daily and the history runs back 20+ years. Filters cover 100+ categories, "from bribery to corruption to forced labor", and the page says repeat news items are kept from raising new alerts. A fintech or crypto firm that screens by API and wants fine control over categories gets the most from it; the trial question is how a 20 year old story weighs against last week's.
ComplyAdvantage
ComplyAdvantage screens adverse media on its own data: its adverse media page says the data "currently screens against over 11,000 verified national, regional, and local media sources across 200+ countries and territories" and is "updated every day". Articles are processed in their native language, and the result is an expert curated profile with images, date of birth, location and article snippets. Monitoring is always on, with name commonness matching to cut noise, and the API can be reconfigured when risk appetite changes. Fintechs and mid sized banks that want curated profiles through one API are the natural buyers. Ask how a source is added or removed, and how often.
Dow Jones Risk & Compliance
Dow Jones sells adverse media two ways. Its data feeds page describes structured adverse media entity data on companies and a separate set of Special Interest Persons involved in legal processes for serious crimes, alongside Factiva news "from its 33,000+ sources in 33 languages across 200 countries". The screening product draws on "thousands of licensed global news sources from Dow Jones Factiva" and names risk categories from regulatory actions to labor issues. Delivery is by feeds, APIs and partner integrations. No refresh interval is stated. Banks that want researched profiles plus licensed news fit best. Ask which hits come from the structured data and which from raw news.
LexisNexis Risk Solutions
LexisNexis screens adverse media in Bridger Insight XG, which lists "thousands of reputable and reliable news media sources across 59 languages" and integrates with WorldCompliance Data, 8M+ curated risk profiles kept by a global research team. Bridger categorizes high probability matches for prioritization and runs in public or private cloud or on premise. A firm can also bring its own data provider into Bridger. Bridger's single engine for account and payment screening, plus the on premise option, suits a bank that keeps screening in house. In a trial, check how news hits and profile hits are told apart in the alert.
LSEG World-Check
World-Check holds adverse media in two layers, as LSEG's adverse media screening page describes them: structured adverse media "researched and organised into World-Check records under defined inclusion criteria", and World-Check One Media Check, an optional add on that searches wider news and web content. Users filter by secondary identifiers, keywords, categories, subcategories and location, with Special Interest Categories for a finer view. Access is through data feeds, APIs and World-Check One. No source count, language count or refresh interval is stated on the page. It fits banks that want researched records first and news second. Ask whether Media Check is in the quote.
Moody's
Moody's adverse media screening page gives "14,000+ Trusted media sources 3+ Billion Articles screened 80+ Risk codes". Sources are reviewed by researchers, foreign language media are monitored in more than 100 countries with English summaries, and keywords, risk codes and thresholds can be aligned with internal policy. Banks and corporates that want reviewed sources and English summaries of foreign press fit it best, provided its risk codes map cleanly to their own offense list.
Owlin
Owlin screens and monitors companies, not individuals, on the pages read. Its screening product covers "over 3 million sources in multiple languages", and its AI agents detect and classify risk signals in context, each linked to its source article. Monitoring runs around the clock with "No batch updates", and the open universe approach covers any business with an online presence, even one missing from company databases. It suits a bank, payments firm or corporate watching suppliers, merchants or portfolio companies. Ask what happens when one story names both a company and its director.
Ripjar
Ripjar's adverse media screening turns more than 6 billion articles, and about 6 million new ones a day, into entity profiles: all media for one entity is grouped into one profile alongside watchlist data, with a risk summary that links every assertion. Profiles are updated twice a day. Natural language processing covers 22+ languages and named entity recognition 50. Large banks whose analysts would rather read one profile per name than a stream of articles are its buyers. A useful trial question: what happens to a profile when a story is corrected or withdrawn?
Sigma360
Sigma360's adverse media screening parses articles from 730,000+ sources in 120+ languages, maps each to named entities and categorizes it by risk type and event topic before an analyst sees it. Materiality scoring escalates events by frequency, severity, source authority and recency, related articles are grouped into one story, and analysts can override any recommendation. Filters and thresholds change without engineering work, and an API is offered. It fits a fintech or bank that wants event level scoring across many languages. Ask to see the materiality score for one known case and the evidence behind it.
Which adverse media screening tool fits your firm?
Picture a payments firm with 4,000 merchants in eight countries, an illustrative case. Half the merchants trade in languages other than English, and most stories about them would surface in local outlets or customer forums. A tool that reads 20 languages and screens raw news by API covers more of that book than a set of curated profiles built on English press. The firm's choice turns on languages, sources beyond news and whether it can screen companies as easily as people. The same reasoning runs through the four situations below.
Which providers offer adverse media screening for a tier one bank?
A tier one bank usually wants researched profiles an examiner recognizes, plus wide news search for EDD. LSEG World-Check, Dow Jones, LexisNexis and Moody's each pair curated records with a news layer; Ripjar builds machine profiles at the scale a large book needs.
Which adverse media monitoring solutions suit a UK bank that screens only at onboarding?
A bank asked by the FCA about ongoing monitoring needs the same names screened every day, not once. The question to each vendor is how monitoring starts for the existing book and how a new story reaches a case. A one off adverse media check shows the gap: it covers the day it runs. ComplyAdvantage, Owlin, Dow Jones and complead describe always on or API triggered monitoring on their pages, and LSEG says its adverse media screening supports "onboarding and ongoing review".
Which adverse media screening tools categorize news by risk type?
All ten sort hits in some way, but the depth differs on the pages read: Castellum.AI lists 100+ categories, Moody's 80+ risk codes, Sigma360 risk type plus event topic, and complead files each hit by the crime described. LSEG and Dow Jones add special interest categories to researched records. Ask for the list in writing, then map it to your own policy.
Which tool screens suppliers, merchants and other companies?
Owlin is built for companies and third parties, and covers businesses missing from registers. Dow Jones offers a structured adverse media set on companies. A firm checking a merchant or supplier also needs each owner and director screened, a check good KYB software runs on every owner it finds.
When is complead not the right fit?
None of the ten publishes a full price on the adverse media pages read; cost usually follows the number of names screened, monitoring frequency, the data sets licensed and whether news search is an add on. Two situations point elsewhere. complead quotes by modules and volume and has no public price list, so a buyer who needs a published price before the first call should look at a vendor that posts one. And complead's adverse media works from settings the firm owns, from source tier to lookback; a team that wants to set none of that and take a vendor's researched profile as it comes will find LSEG World-Check or Dow Jones closer to what it wants.
What changed in adverse media screening in 2026?
- 1 June 2026. Ripjar published a survey of 400 financial services decision makers in the UK, US, France and Germany, with fieldwork in April 2026. According to Ripjar, 77 percent conduct adverse media screening and 58 percent still rely on manual internet searches.
- 3 August 2026. FinCEN's UBS penalty named unread and unexplained negative news as a customer due diligence failure, and the order sends an independent reviewer to check how media reports reach each risk profile.
- 28 September 2026. World-Check moved onto Active Intelligence, an LSEG engine built on Amazon Bedrock, so a World-Check trial this year tests that new engine rather than the old one.
Whichever names make the shortlist, the decision underneath is the same: researched profiles that a vendor maintains, or raw news screened against filters the firm sets and defends. The first is narrower and easier to explain; the second reaches further and needs a policy behind every setting. Put last quarter's hits through two tools of different kinds and compare what each would have kept, and why.
