Same name, different person
A search engine doesn't know your customer's date of birth or city. Investigators do that matching by hand, and two investigators can reach different answers on the same result.



28 specialized agentsAll systems operational
Ready to transform your security infrastructure?
Explore our complete agent library and request a custom demoView All Solutions
Aiden FluxSenior AI Fraud Risk AnalystFraud Detection & Risk Scoring
Rhea LedgerSenior AI KYC/AML Compliance DirectorKYC/AML & Sanctions Screening
Nova SentinelLead AI Zero Trust Security ArchitectZero Trust Access Security
Iris VermaAI Verification SpecialistIdentity Verification & KYC
Oscar GraySenior AI OSINT Intelligence DirectorOSINT & Threat Intelligence
Bella NovaAI BNPL Risk AnalystBNPL Risk Monitoring


28 specialized agentsAll systems operational
Ready to transform your security infrastructure?
Explore our complete agent library and request a custom demoView All Solutions


28 specialized agentsAll systems operational
Ready to transform your security infrastructure?
Explore our complete agent library and request a custom demoView All Solutions
Oscar Gray — Senior AI OSINT Intelligence DirectorOscar Gray is an AI agent that searches adverse media, breach databases, dark-web mentions and domain and IP reputation for the customers and cases your team is reviewing. He drops namesakes with a recorded reason and attaches what's left with a confidence level. Your analyst reads the evidence and makes the call.

An adverse media search on a common name returns pages of results about other people. Investigators read them one by one, copy links into the case and write down why each one doesn't fit. The relevant article, when it exists, is often on page four.
of hits on a common name
Most are about someone else.
A search engine doesn't know your customer's date of birth or city. Investigators do that matching by hand, and two investigators can reach different answers on the same result.
News archives, breach data, dark-web mentions and domain reputation sit in different tools. Pulling them together for one case takes a stack of browser tabs.
Examiners ask what was searched, when, and why a result was set aside. A pasted link in a case note doesn't answer that.
Oscar Gray is a Senior AI OSINT Intelligence Director. He runs outside-in checks on customers, counterparties and cases, filters out what doesn't fit and attaches what does to the file your analyst reviews.

We don't publish relevance scores from our own tests. Run Oscar beside your investigators and measure what he finds, and drops, on your own customers.
Oscar Gray connects beside your systems through APIs. Your case management stays where it is.
A search request arrives from onboarding, screening, a fraud alert or an analyst. It carries the identifiers you hold: name, date of birth, nationality, address, company number, email or domain.
Oscar queries the sources you license and allow, such as news archives, breach data providers, dark-web monitoring feeds and domain and IP reputation services.
Each result is matched against the subject's identifiers. Namesakes drop out with a recorded reason. What's left gets a confidence level and a short summary of why it fits.
Findings go into the case with source, date and confidence. Your autonomy settings decide routing: medium and high always reach an analyst. The search, its sources and every discard go into tamper-evident evidence storage.
Run Oscar Gray in shadow mode on a sample of your customers and open cases. He searches, filters and attaches findings, and nothing is closed. Compare his results with your investigators' before you switch anything on.
Oscar doesn't make you compliant. He produces the evidence these frameworks expect you to keep.
Fewer pages of search results. The findings that fit, with their sources.
| CRITERIA | Manual web searches | Adverse media screening tool | Oscar Gray |
|---|---|---|---|
| Time to first results | Depends on the investigator's queue | Vendor setup and threshold tuning | Shadow mode on your live cases |
| Who decides | Investigator | Match threshold, then analyst | Analyst, inside autonomy bands you set |
| Namesake handling | By hand, varies by person | Name and keyword matching | Matched on your identifiers, with the reason recorded |
| Sources covered | Whatever the investigator thinks to check | Media only, in most tools | Media, breach data, dark-web mentions and domain reputation you license |
| Record of the search | Case notes and links | Hit list | Full log of sources, results and discards |
| Where it's weaker | Slow and inconsistent at volume | Noisy on common names | Only sees the sources you license; nothing unpublished or unindexed |
Oscar's findings matter most when they land inside a case another agent is already building.

Uses Oscar's findings in screening dispositions and customer risk ratings.
Meet Rhea
Asks Oscar for breach data and adverse media when a fraud alert needs outside evidence.
Meet Aiden
Pulls public records on claimants and repairers before an SIU referral.
Meet ClaraLow risk can run on its own if you allow it. Medium risk goes to a person by default. High risk always goes to a person. You set the bands per rule, channel and transaction type.
Turn Oscar off without touching the other agents or your core systems. The switch, and who used it, is stamped on the record.
Run Oscar on live data with nothing blocked or closed. Compare the calls with your team's before anything changes.
Every decision answers why, in plain English, with the signals and the rule or policy behind it.
Each decision is stored with its inputs, its reasoning and the person who approved it, in tamper-evident evidence storage.
Agents connect beside your systems through APIs. Your core banking, screening and case tools stay where they are.
What we're learning about AML, fraud and the evidence examiners ask for.






Talk to the people who build the agents. We'll answer per capability, yes or no.
An AI agent that runs open-source intelligence checks for compliance and fraud teams. Oscar Gray searches adverse media, breach databases, dark-web mentions and domain reputation, filters out namesakes and attaches relevant findings to the case with a confidence level. Your analyst reviews them.
He compares each result with the identifiers you hold, such as date of birth, city, nationality, employer or company number. Results that conflict drop out with a recorded reason. Results he can't confirm either way go to an analyst.
The ones you license and allow. Typical sources are news archives, breach data providers, dark-web monitoring feeds and domain and IP reputation services. We agree the source list with you during deployment design.
Your analyst does. Oscar can close results that are clearly about someone else, only where you allow it. Anything relevant goes to a person. A kill switch turns Oscar off without touching your other systems.
Public data about a person is still personal data under GDPR, UK GDPR and similar laws. Oscar searches only for the purposes and sources you configure, and logs every search so your data protection team can review it.
Oscar searches a sample of your customers and open cases, but nothing is closed. Your investigators keep working as they do today, and you compare Oscar's findings with theirs. You decide whether, and where, to switch on any autonomy afterwards.
FluxForce runs as SaaS, on-premise or hybrid, built on Microsoft Azure. We agree data residency and which components run inside your environment during deployment design, before any data moves.
Run Oscar Gray beside your current process. He works on your live data and records every call, and nothing is blocked, closed or sent until you decide.
Shadow mode results belong to you.
Start with one workflow in shadow mode, then decide how much each agent does on its own.