Methodology
Understand the evidence behind the verdict.
AdverseMe searches public and integrated sources, checks which results concern the supplied person or company, and uses four independent AI judges to assess retained findings. The screening record keeps source outcomes and limitations so your team can review the result.
Product explanation reviewed · AdverseMe
What sources does AdverseMe search?
The coverage directory lists 110+ integrated sources, including sanctions, regulators, registries, courts and news. 13 official feeds are scheduled twice daily; 5 GCC and MENA national lists are scheduled daily. A catalogue entry describes integration coverage, not a guarantee that every source returns data on every run.
Some sources apply only to particular countries or entity types. Upstream outages, access restrictions and missing identifiers can limit a search. The Engine Trace records outcomes such as results returned, no results, skipped and unavailable. A missing result and a failed search are different states.
How are namesakes separated from the subject?
A name is a starting point. Identity review uses supplied identifiers and source evidence such as nationality, date of birth, location, employer and company details. Arabic transliteration handling helps compare spelling variants. Wrong-entity articles are excluded from adverse findings and retained with a reason where the identity gate records one.
Adverse media screening concerns relevant negative information about the subject. A routine mention, a syndicated duplicate and an allegation are not interchangeable evidence. Review the original article, its date, source and the subject's role in the event. The Wolfsberg Group's Negative News Screening FAQs provide industry guidance on designing a risk-based process.
How is the risk score produced?
The AI council assesses retained findings independently. The result combines the available judges' scores using the median and records disagreement. Judges can be unavailable; an unscored result is labelled accordingly. Four models can still share errors, so the source evidence and a human review remain important.
Temperature zero and fixed seeds reduce sampling variation. AdverseMe also stores evidence signatures and reuses matching verdicts within a 14-day window. Reuse depends on the subject, relevant evidence, list versions and methodology remaining unchanged. A changed input can produce a new result.
How should ownership findings be read?
The ownership graph separates registry evidence from reported or inferred relationships, preserves percentage bands and identifies unknown stakes. These findings support review of potential sanctions exposure; a graph may be incomplete when records are unavailable.
For US sanctions, consult OFAC's indirect ownership examples and its distinction between ownership and control. An analyst should review the relationships against the rules that apply to the transaction.
How do we publish product and regulatory information?
AdverseMe publishes its product documentation and blog under the company name. Product explanations are checked against the implemented workflow. Our shared product catalogue supplies the source counts and subscription prices shown across the website.
Regulatory explainers link to issuing authorities and should be read with their current guidance. Publication and review dates are shown on articles; a date change follows an actual revision. Examples are labelled, and screenshots identify the product screen they show.
Send corrections with the page URL and supporting evidence through our contact page. For processing information, review the privacy policy and sub-processor list.
Frequently asked questions
Does a screening result make the compliance decision?
No. AdverseMe presents evidence and an AI-assisted assessment. Your team confirms the identity, reviews relevant sources and decides what action is required for the relationship and applicable rules.
Does no match mean there is no risk?
No match means the completed searches did not return a retained match for the supplied subject details. Coverage, unavailable sources, identity uncertainty and new information can affect the result. Review the trace and limitations before relying on it.
Where can I see the AI judges' reasoning?
The screening result in the app shows the council's votes and reasoning. PDF reports present the determination, findings, sources, coverage, excluded matches and limitations; they do not include the per-judge table.