Deconflict

Pattern recognition surfaces overlapping cases, shared adversary infrastructure and common victim wallets across separate agencies, correlation no analyst reaches by hand. Strip it out and a credentialed intelligence exchange between agencies and regulated institutions still works, and that network is the asset.

Scoring, ranking and likelihood estimates are explicitly not offered; verified matches go to analysts, and the institution owns the decision a signal informs. Intervention before a transaction settles is the central promise, yet where a person sits in that seconds long sequence, and what happens when a match is wrong, is not described.

Matches are verified rather than scored, and each arrives with its typology, so the model risk sits in the matching itself. False positive rates and attribution accuracy are unpublished.

More than 2,000 law enforcement agencies are connected, with the Secret Service, the FBI and the Los Angeles Police Department among those named. No bank, credit union or exchange is named on the institution side, and no outcome is quantified.

Training is ruled out without a carve out: personal information and wallet intelligence submitted by users never train, retrain or fine tune any model. Participation needs verified credentials, investigative files stay with the agency and wallet intelligence is kept indefinitely, though supplier retention terms are missing.

Identifiers are exchanged in hashed form, so institutions get investigative context on wallets without personal information exposure. The privacy policy updated May 1, 2025 sets retention and gives registered users access, correction and deletion, while the subjects of signals have no rights route.

SOC 2 Type II is named in the privacy policy with AES 256 at rest and TLS 1.3 in transit, and an AICPA badge appears on the homepage. Assessment against the Criminal Justice Information Services security policy, which binds participating agencies, is not mentioned.

Output is shaped for alert investigation, escalation decisions and suspicious activity report narratives, with supervisors named as the review standard it must withstand. Naming a supervisor is not a formal admission such as a register entry, and no license is held or needed.

A person investigated but never charged can have payments stopped or an account closed on a signal they will never see or contest. Law enforcement attention falls unevenly across places and populations, so a network of agencies aggregates those patterns, and no outcome or false attribution analysis is published.

An audit trail and platform logging of attribution and access let a compliance officer justify or reverse an account closure. Subjects of signals have nothing, since the basis is an undisclosable law enforcement matter, and Deconflict publishes no guarantee or indemnity.

Each response names the originating agency, so an institution knows which body produced a finding rather than receiving an anonymous flag. AWS is the named host, while processors appear only by category.

Delivery runs through an API, dedicated environments and Model Context Protocol support, with webhook alerts on monitored wallets. Transaction monitoring, case management and core platforms on the institution side are unnamed.

AWS GovCloud and dedicated environments are offered, with data stored on AWS infrastructure in the US. Where international agencies' signals rest, and under whose jurisdiction, is not addressed.

Neither agencies nor regulated institutions see published pricing, and the two sides are unlikely to pay on the same basis. Whether institutions pay per query, per monitored counterparty or by subscription is left open.

Banks, credit unions and exchanges on the regulated side meet federal, tribal and international agencies on the other, each side's value depending on the other's presence. Typologies run from money mule networks and investment scams to business email compromise, centered on digital asset investigations.