Tech Support : support@safience.io

Known-offender identification at the bank branch: what it is and what it is not

A bank will spend weeks learning who a loan applicant is. It will check identity, credit, sanctions lists and account history before it lends a dollar. Then it opens the branch doors at nine and knows almost nothing about the people who walk up to the teller line.

That gap is not theoretical. The FBI’s most recent Bank Crime Statistics report counts 1,263 robberies of financial institutions in 2023. Firearms were used in 230 of them and weapons were threatened in another 539. Twenty-one people were taken hostage. And robbery is only the loudest branch crime. Federal regulators report that suspicious activity reports for check fraud nearly doubled between 2021 and 2023, and much of that paper still crosses a counter in person.

Fraud teams watch the transaction. Almost nobody watches for the known offender before the transaction starts.

What known-offender identification means

Known-offender identification answers one narrow question: is someone who matches a verified law-enforcement record, relevant to this branch, on the premises right now?

It is not a background check on every customer. It is not a watchlist the bank builds from its own suspicions. And it is not surveillance of the public. The comparison set is a defined list of records that a court or agency has already created: active felony warrants, fugitives, registered sex offenders where relevant, and missing persons a branch employee might help find.

How it works at a branch

With RTIS, the Real-Time Threat Identification System, a branch uses the cameras it already has at the entrance.

  1. Capture. A single small still image is taken at the door. There is no video stream to Safience and nothing is recorded.
  2. Compare. The image is checked against relevant records in UMbRA, the law-enforcement-sourced identity database, which holds 58.4M+ records from law-enforcement sources only.
  3. Discard. If there is no match, the image is deleted at once. The bank never learns who walked in, and no biometric template is stored.
  4. Review. A potential match goes to a trained analyst, who checks the record, its currency and the context before anything is sent.
  5. Alert. Only a verified, relevant match becomes an alert to the branch’s designated contact, typically in under a minute, with guidance on what to do next.

Most customers are never part of the process at all beyond step three. That is the design.

What it is not

It is not a way to refuse service to people with old convictions. Relevance rules decide which records matter at a bank, and a decade-old misdemeanor is not one of them.

It is not automated. No alert reaches the branch without a human decision, and the branch decides what to do with it under its own policy.

It is not a credit tool or a KYC substitute. Account onboarding has its own legal framework. Branch identification is about physical safety on the premises.

Where it helps most

The teller line. A person with an active warrant for armed robbery walking into a branch is precisely the alert a branch manager wants before that person reaches the counter, not after.

Repeat check presenters. A fraud offender with an active warrant may try more than one branch in a day. A verified alert at the second branch can stop the third.

Employees at risk. Bank staff are sometimes followed to work by someone with a history of violence toward them. A person with an active warrant tied to a domestic-violence case is a legitimate, specific alert.

Insider risk. Customers are only half the picture. eMotive continuous, consent-based employee monitoring tells a bank, with the employee’s written consent, when a team member is newly arrested, rather than waiting for the next periodic background check.

Questions to ask any provider

  1. Is every record in the comparison set sourced from law enforcement, and can you name the sources?
  2. What happens to an image that does not match, and how quickly?
  3. Do you keep video or biometric templates? (The right answer is no.)
  4. Who reviews a match before the branch is alerted?
  5. How do your relevance rules work, and can we see them?
  6. How do you handle biometric privacy law in each state where we have branches?

That last question matters. Illinois, Texas, Washington and a growing number of cities set their own rules on biometric information. Our state privacy policy pages explain how Safience operates in each.

The short version

Banks already accept that knowing who they are dealing with is part of managing risk. Known-offender identification extends that idea from the loan file to the lobby, using records that already exist, a human in the loop, and nothing kept on anyone who is not a match.

See how it fits your branches on our Safience for banks page, or request a Threat Mapping Assessment.

Frequently asked questions

Does Safience record video in bank branches?

No. Safience works from a single small still image and never receives, records or stores video.

What happens to images of customers who are not a match?

They are deleted immediately. No biometric template is created or kept.

Who decides whether a match is real?

A trained Safience analyst reviews every potential match before any alert reaches the branch. The branch then acts under its own policy.

Which records are used?

Only verified law-enforcement records relevant to a bank setting, such as active felony warrants and fugitives, drawn from UMbRA, the law-enforcement-sourced identity database.

Sources

  • FBI, Bank Crime Statistics 2023. https://www.fbi.gov/file-repository/reports-and-publications/bank-crime-statistics-2023-091724.pdf/view
  • Consumer Compliance Outlook (Federal Reserve), “Responding to Counterfeit Instrument Scams and Mail-Related Check Fraud,” 2025, citing FinCEN. https://www.consumercomplianceoutlook.org/articles/2025/second-issue-2025/responding-to-counterfeit-instrument-scams-and-mail-related-check-fraud

Featured photo by Anton Acosta on Unsplash.

Previous Next
Other Articles