Historical edition. This document does not describe the current service configuration.
Version 2026-09-24.2 · Permanent link to this version · U.S. legal documents
Version history
- Version 2026-09-24.2 — Effective 2026-09-24 · Earlier published version
1. Information for organizations using Evidize
Evidize LLC supplies application analysis and interview-access screening software for U.S. private employers and staffing agencies. This document explains the service and its limits. It is not an employer's candidate notice, a new contract or confirmation that a particular hiring process meets the law.
The organization selects the job criteria and makes hiring decisions. Evidize supplies its own processing information and the assistance required for its actual role. See the state guide for the conditions that apply to your use.
2. What the service does
Configured application analysis extracts information from submitted résumés, organizes qualifications and compares available evidence with the organization's job criteria. It can produce summaries, assessments, scores, recommendations and rankings. Approved analysis and ranking run automatically.
Interview screening checks permitted application, network, clipboard-activity and display indicators. The actual checks are explained to the participant for the session. A security hold is temporary; a valid clear check can release meeting access automatically. A security signal is not proof of cheating or a hiring rejection.
Do not use the service for unlawful discrimination, undisclosed surveillance, unrelated candidate profiling or unsupported government or government-contractor hiring. This service description does not add biometric identification, recorded-video analysis, credit checks or criminal-history screening.
3. Inputs, outputs and configuration
| Function | Information used and result |
|---|---|
| Résumé extraction | Submitted résumé text is organized into profile fields, missing information and supporting evidence. The extraction prompt identifier is candidate_profile_extraction:v1. |
| Qualification assessment | Extracted profile information supports a structured qualification assessment using the baseline_v2 rubric. An unsupported qualification remains unresolved. |
| Job fit | Profile information and the organization's requisition criteria produce criterion findings and fit indicators. The structured assessment uses candidate_job_fit_assessment_v1 with the job_fit_v2 rubric. Configured criteria and weights affect the result. |
| Screening | The Windows/macOS collector description is identified as windows-macos-2026-09-13.1. Session settings and available operating-system permissions determine the checks. Uploaded clipboard events contain activity information, not clipboard text or copied file names. |
| Hosted AI configuration | At the September 24, 2026 configuration check, the production Azure Foundry account had no model deployment and hosted AI execution was disabled. This edition does not identify or approve an active model for that configuration. |
These identifiers describe the reviewed implementation. A changed model, prompt, purpose or configuration needs corresponding release information. This document does not establish a model's training-data provenance, a performance benchmark or validation for every job.
4. Limits and appropriate use
Extraction can misread a document. An assessment can omit evidence, misunderstand experience or reflect bias. The information in an application may be incomplete or inaccurate. A score is an assessment against configured criteria, not a probability of successful employment. Missing evidence alone is not a confirmed failure.
Screening results depend on the operating system, permissions, connectivity and the tools detectable by the enabled checks. A clear check is not a guarantee that no assistance was used. A flag needs context, including approved assistive technology and corrections supplied by the participant.
Check the source material and job criteria before relying on an output. Investigate a disputed or inconsistent result through the existing support process. Provide the accommodations and review required for your hiring use. Evidize does not provide an independent bias-audit result or claim that these checks establish validity for every employer's selection procedure.
5. Provider disclosures and updates
For covered developer activity under Colorado section 6-1-1702 from January 1, 2027, Evidize must supply the applicable product information and directly notify affected organizations of qualifying changes within a reasonable time. The covered developer records must be retained for at least three years from creation. A general product page does not replace release-specific information required for a covered deployment.
For covered Connecticut deployments from October 1, 2027, Public Act 26-15 requires applicable provider information supporting the employer's disclosures. The employer supplies its own purpose, decision process and contacts. Where information is lawfully withheld, Evidize must provide the explanation required by the applicable law.
Direct provider-information updates use email and do not request contractual acceptance. Separate permission or a contract change follows its own applicable process. Contact privacy@evidize.com or your existing support contact for information or technical assistance.
6. Training-data information
California's training-data disclosure law has separate coverage and content requirements. If it applies to an Evidize release, Evidize must publish the factual dataset description before the covered release. An Azure model label, this service description or an employer's notice cannot substitute for that description. A security-only exception does not automatically cover hiring analysis.
The Privacy Notice describes how submitted information is used. Customer-data use is governed by the Customer Data Agreement and the applicable supplier arrangements. This document grants no additional permission to train on candidate information.