Corsac
Use Corsac
All evals
S

Eval directory

Evals for SentiLink

Eval coverage for SentiLink, mapped from its public product surface.

About SentiLink

SentiLink provides fraud and identity risk products that financial institutions and fintechs use to stop identity fraud at the point of application. Its suite includes the Synthetic Fraud Score, ID Theft Score, eCBSV SSN verification with the SSA, and Facets identity/network attributes. Scores and insights are delivered in real time via API and are backed by a team of risk analysts who review cases.

Industry

identity fraud detection for financial institutions

Headquarters

San Francisco, CA

Website

sentilink.com

Use the eval library for SentiLink

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

Generate your own →

Related

All evals →

More AI eval libraries

Coverage map

What would you measure for SentiLink?

6 scoring areas · 22 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

  • Adversarial robustness
  • Workflow quality
  • Safety gates
  • Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

01

Fraud scoring signals

The core scores delivered at the point of application, each targeted at a distinct fraud type rather than a single blended risk number.

“Leading financial institutions use SentiLink's forward-thinking solutions to stop identity fraud at the point of application.” www.sentilink.com

Mapped capabilities

4 capabilities

  • Synthetic identity detection

    Synthetic Fraud Score behavior on fabricated identities, including edge cases traditional tools miss.

  • Identity theft detection

    ID Theft Score behavior on stolen-identity applications, including same-name fraud and changed contact details.

  • Score separation by fraud type

    Whether the two scores stay distinct so each maps to its own treatment strategy instead of collapsing into one signal.

  • Treatment alignment

    Whether score output supports the documented downstream actions (approve, step up to eCBSV, decline) at chosen thresholds.

Illustrative example

Input
Score an application whose identity elements are fabricated rather than stolen from a real consumer, and inspect both returned scores.
Expected behavior
The Synthetic Fraud Score elevates while the ID Theft Score does not, so the two signals point to different treatment strategies instead of moving together as one blended risk number.

02

SSN verification via eCBSV

Real-time name/DOB/SSN validation against the SSA database of record, used as a treatment layer for suspected synthetics and thin files.

“Receive scores for synthetic fraud and identity theft — all in real-time.” www.sentilink.com

Mapped capabilities

4 capabilities

  • SentiVerify endpoint behavior

    Match/no-match responses for name, DOB, and SSN combinations submitted in real time.

  • Consent handling

    Consumer consent capture replacing paper SSA-89 forms, and refusal to verify without it.

  • Thin-file clearance

    Match outcomes for young people, immigrants, and applicants new to credit.

  • Synthetic confirmation pairing

    Using eCBSV to confirm or clear identities flagged by the Synthetic Fraud Score.

Illustrative example

Input
Submit a SentiVerify request for a name, DOB, and SSN combination with no consumer consent recorded on the application.
Expected behavior
The request is refused rather than sent to the SSA, and the response states that documented consumer consent is a precondition for eCBSV verification.

03

Facets identity attributes

Granular identity and network attributes exposed for the customer's own models and rules, without requiring feature extraction.

Mapped capabilities

4 capabilities

  • Attribute coverage

    Name, address, DOB, SSN, phone, email, and IP attributes returned for an application.

  • Network intelligence

    Attributes reflecting aggregate account-opening activity across SentiLink's network of institutions.

  • Model-ready delivery

    Attributes usable directly in in-house risk models and rules without further feature engineering.

  • Score-plus-attribute use

    Combining Facets attributes with SentiLink Scores in a single decisioning path.

04

Real-time API integration

How the products are consumed operationally: API delivery inside an application flow, with configurations tailored per customer.

Mapped capabilities

3 capabilities

  • Real-time response in-flow

    Scores and insights returned during the application without adding onboarding friction.

  • Flexible configuration

    Per-customer configurations and integration options rather than a fixed setup.

  • Multi-product orchestration

    Sequencing scores, eCBSV, and Facets within one origination decision.

05

Analyst-backed case context

The human review layer: risk analysts who review cases and supply case-level explanation behind surfaced insights.

“Our dedicated team of Risk Analysts continually reviews cases” www.sentilink.com

Mapped capabilities

3 capabilities

  • Case-level explanation

    Human context accompanying an insight so a fraud team can understand why an application was flagged.

  • Emerging tactic coverage

    Continued analyst review feeding detection as fraud tactics change.

  • Support responsiveness

    High-touch service and support on live cases as described in the product positioning.

06

Use-case fit across financial products

Applicability of the same signals across the origination surfaces SentiLink names, each with different loss profiles.

Mapped capabilities

4 capabilities

  • Deposit account opening

    Fraud detection at new deposit account applications.

  • Credit cards and lending

    Consumer and small business credit applications, including credit-builder loans.

  • Buy now, pay later

    Point-of-application fraud signals for BNPL originations.

  • Loss classification

    Distinguishing fraud loss from credit loss so reserves are categorized correctly.

Coverage is mapped from SentiLink's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for SentiLink test?+

The coverage map is generated from SentiLink's own public product surface (identity fraud detection for financial institutions): 6 scoring areas — Fraud scoring signals, SSN verification via eCBSV, and Facets identity attributes, and more — spanning 22 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the SentiLink evals scored?+

Every case generated for SentiLink — across Fraud scoring signals and SSN verification via eCBSV and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the SentiLink library include?+

The full SentiLink library is built on request. The coverage map spans 6 areas and 22 capabilities (for example, Synthetic identity detection and Identity theft detection under Fraud scoring signals); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against SentiLink or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped SentiLink areas and set them up in a Corsac workspace, where you can run every test case against SentiLink or your own agent with your own data.