AI Governance Statement

Effective Date: September 8, 2026 -- Version 1.1

How Zenna uses artificial intelligence in hiring and candidate outreach. Our starting principle is simple: technology can help people make better hiring decisions, but it should not remove accountability, hide the evidence, or quietly exclude qualified people.

1. Purpose and scope

Zenna is the human-guided hiring platform developed by Rokstars Global HRM Inc. This statement explains how Zenna uses artificial intelligence, automated technology, and human review when supporting employers and candidates.

It applies to Zenna features that currently help create and organize hiring information, parse resumes, compare candidate evidence with role criteria, produce fit guidance, and support hiring workflows. It also describes governance requirements for planned features, including Rolodex matching and automated, consent-based candidate outreach. Those planned features are not currently active.

This statement is intended to be read together with Zenna's Privacy Policy, Accessibility Statement, Terms of Service, and any notices shown during a particular application or outreach process.

2. Our commitments

People remain accountable for consequential hiring decisions.
AI output is guidance, not unquestionable fact.
Candidates are assessed against role-related criteria established for the specific opening.
Accommodation information is handled separately from qualification scoring.
Candidates are told when AI is used where notice is required or appropriate.
Candidates can ask questions, correct material inaccuracies, and request human reconsideration.
Opt-in candidate outreach is relevant, limited, transparent, and easy to stop.
Changes to models, prompts, scored inputs, and decision workflows require review and testing before production use.
When Zenna cannot evaluate information reliably, the issue is sent to a person rather than treated as evidence against the candidate.

3. Where Zenna uses AI and automation

AI and automation may support the activities below. The exact features used can depend on the employer, role, service plan, and jurisdiction.

Currently active uses

Activity	What technology may do	What it does not decide alone
Resume processing	Extract and organize work-related information from a resume.	Whether the person is hired or rejected.
Candidate fit guidance	Compare permitted evidence with role criteria and produce a score, label, and plain-language rationale.	The final hiring decision. Human review is required before a candidate recommendation reaches an employer.
Screening responses	Assess whether an answer responds to the question and provides relevant evidence.	Writing quality, disability, language background, or accommodation need.
Workflow support	Support reminders, status updates, scheduling, and organization of information.	A consequential employment decision without accountable human involvement.

Planned features not currently active

Zenna may later use automation to identify opted-in candidates whose stored qualifications and preferences relate to an opening and to prepare or send limited messages about potentially relevant opportunities. These features will not be activated until consent, suppression, message-frequency, accessibility, privacy, technical, and legal controls have been implemented and verified. A potential match will not be an offer, ranking, guarantee, or hiring decision, and silence or refusal will not count against a person.

4. Human responsibility and review

Zenna is designed as a human-guided service. A numerical score is not a hiring decision. It is one piece of guidance that must be considered with the underlying evidence, the role criteria, and human judgment.

Before a candidate recommendation reaches an employer through Zenna's managed process, an authorized human reviewer or advisor examines the available evidence. Reviewers can question the output, add context, request clarification, and override a recommendation. Material overrides should include a reason.

Zenna is designed so that a score does not replace human review or an employer's lawful, job-related decision process. Employers remain responsible for their decisions, including activity outside Zenna, and for complying with the laws that apply to them.

5. How candidate fit guidance works

Zenna compares candidate evidence with the requirements and success measures established for a particular role. The scoring configuration may consider permitted information such as:

Skills and qualifications connected to the role.
Relevant experience against the role's stated requirements.
Work arrangement compatibility, such as onsite, hybrid, or remote availability.
Relevant evidence contained in screening responses.
Permitted work-related information extracted from a resume.

Where a screening answer is incomplete or does not respond to the question, Zenna may identify that the requested evidence was not provided. It must not apply an arbitrary punishment. The issue is named for a human reviewer, who can consider misunderstanding, accessibility barriers, technical problems, or other context.

Zenna does not treat spelling, grammar, punctuation, writing style, brevity, or English-language fluency as evidence of job fit unless a specific communication requirement is lawful, necessary for the role, disclosed in advance, and assessed through an approved method.

6. Information excluded from qualification scoring

Zenna's scoring instructions exclude the following from qualification scoring:

Accommodation requests and accommodation-management notes.
Disability or medical information.
Candidate names.
School names and graduation years.
Employment gaps without relevant context.
The job board, source, or channel through which the candidate arrived.
Location beyond a legitimate work-arrangement or legal-work requirement.
Writing fluency, grammar, and language polish in screening responses.

Some excluded information may still appear in source documents such as a resume. Zenna's governance program requires technical controls that minimize what reaches the scoring model and testing that checks whether prohibited information influenced the result. Until those controls are fully verified, Zenna describes the current safeguard precisely: the model is instructed not to use the excluded information, and a human reviewer examines the recommendation.

7. Accommodation requests

Candidates may request an accommodation within the hiring process. Accommodation information is stored separately from qualification scoring information and is assigned to a Rokstars advisor for follow-up.

The scorer does not query Zenna's accommodation-request records. An accommodation request must not raise or lower a fit score, be used as evidence about capability, or disappear because an automated process could not handle it.

Employers do not receive a diagnosis or unnecessary personal detail through Zenna. When employer involvement is needed, the advisor communicates the functional accommodation need and appropriate process information, subject to the candidate's direction and applicable law.

8. The Zenna Rolodex and candidate matching

The existing Zenna Rolodex records candidates who have opted in to be considered for relevant future opportunities. Persistent candidate profiles, automated matching, and automated re-engagement are planned features and are not currently active.

Before any automated Rolodex matching or outreach is activated, the following requirements must be implemented, tested, and approved:

A person must have provided valid consent or another lawful basis must clearly permit the contact.
Zenna uses work-related qualifications, experience, role interests, location or work-arrangement preferences, and other permitted matching information.
The match means only that an opening may be relevant. It is not an offer, guarantee, ranking, or hiring decision.
The message identifies Zenna or Rokstars, explains why the person is being contacted, and provides enough information to decide whether to learn more.
Every automated outreach message provides a clear way to stop future opportunity messages.
Withdrawing consent, declining an opening, or ignoring a message does not lower a score or reduce the person's standing in another application.
Zenna monitors message frequency and relevance to avoid repetitive or misleading outreach.
Consent preferences and opt-out requests are recorded and honoured promptly.
Candidate records are not repurposed for unrelated advertising or sold as marketing lists.

9. Candidate notice, choice, and recourse

Zenna aims to give candidates understandable information and a practical way to reach a person. Depending on the feature and jurisdiction, candidates may:

Receive notice that AI or automated tools are used in the hiring process.
Ask what information was used to support a recommendation.
Request correction of materially inaccurate candidate information.
Ask for human reconsideration when they believe relevant evidence was missed or misunderstood.
Request an accommodation without affecting qualification scoring.
Withdraw from Rolodex outreach or change communication preferences.
Request access to or deletion of personal information, subject to legal, contractual, security, and record-retention requirements.
Report an accessibility, privacy, or fairness concern to a person at Rokstars.

An explanation may describe the role criteria and evidence considered without revealing another candidate's information, confidential security controls, or protected employer information.

10. Data protection and service providers

Zenna uses carefully selected service providers to support parts of the platform, including AI-assisted candidate fit guidance, resume parsing, communications, scheduling, and platform infrastructure. These providers process information only for defined business purposes and according to applicable contractual, privacy, and security requirements.

Zenna does not intentionally use candidate application information to train general-purpose AI models. Before approving a provider for production use, Rokstars reviews its data-use terms, retention practices, security commitments, processing locations, incident procedures, and available controls.

Access to candidate information is limited according to role and operational need. Accommodation information is kept separate from qualification scoring and is subject to additional access controls. Retention periods are described in Zenna's Privacy Policy and supporting retention practices.

Service providers may change as Zenna's technology develops. Any provider handling candidate or employer information must complete an appropriate privacy, security, technical, and legal review before production use.

11. Testing, change control, and monitoring

AI models and legal requirements change. Zenna therefore governs an approved scoring release, not merely a provider name. A scoring release includes the model, prompt, role criteria, parser, permitted inputs, knowledge-base instructions, and related workflow controls.

Before a material change reaches real hiring decisions, Zenna's governance process requires:

A documented description of the proposed change and its purpose.
Testing with fictional or properly de-identified cases, including accessibility and non-responsive-answer scenarios.
Checks that accommodation information and other prohibited inputs do not influence scoring.
Review of score changes, rationales, parsing failures, cost, reliability, and human disagreement.
Approval by an accountable person with appropriate technical or employment expertise.
A rollback plan when a change materially affects scoring or candidate treatment.

Production monitoring should include scoring failures, unexplained score-distribution changes, missing rationales, reviewer overrides, parsing problems, accommodation-response ownership, outreach opt-outs, and complaints. Fairness or outcome analysis involving sensitive demographic information requires separate lawful collection, access controls, and legal review.

12. Records and explanations

Zenna is building toward a contemporaneous record for each score that identifies the approved scoring release, evidence considered, score and rationale, time produced, human review, and any override. Where possible, records should use references and tamper-evident hashes rather than unnecessary copies of resumes or sensitive information.

Until the full provenance system is implemented and verified, Zenna must not claim that every score has a complete technical receipt. Existing score, rationale, application, role, and reviewer records remain part of the available decision record.

13. Security and incident response

Zenna applies access controls, logging, separation of sensitive information, and secure credential handling to reduce unauthorized access or misuse. Security controls are reviewed as the platform changes.

If Zenna identifies a material scoring, privacy, security, or outreach problem, Rokstars may pause the affected automation, preserve relevant records, investigate the scope, move affected cases to human review, correct the issue, and provide notices where required. Resuming the feature requires documented approval and appropriate testing.

14. Employer responsibilities

Employers using Zenna are expected to:

Provide accurate, job-related requirements and avoid criteria that unlawfully exclude candidates.
Review the evidence and rationale rather than relying only on a score or label.
Keep a person accountable for the final hiring decision.
Provide required notices about AI use in job postings and hiring processes.
Participate in accommodation processes and protect candidate confidentiality.
Use Zenna information only for the authorized hiring purpose.
Report suspected errors, bias, security issues, or misuse promptly.

15. Accountability and review

Rokstars Global HRM Inc. is accountable for Zenna's governance. Product, HR, technical, privacy, accessibility, and legal expertise are brought into decisions according to the level of risk. Material changes, new jurisdictions, new scored inputs, new providers, and any move toward automatic rejection require heightened review.

This statement will be reviewed at least annually and sooner when Zenna materially changes its AI use, introduces a new consequential feature, enters a new jurisdiction, or identifies a legal or operational change that affects the commitments described here.

16. Limitations

AI can be incomplete, inconsistent, or wrong. A fit score does not measure a person's worth, potential, character, or complete ability to succeed. It reflects only the permitted evidence available for one role at one point in time.

Zenna does not promise that AI is perfectly unbiased or that one governance statement satisfies every law in every jurisdiction. Rokstars commits to reasonable testing, transparency, human accountability, correction, and continuous improvement as the technology and legal landscape change.

17. Contact us

Questions, correction requests, accommodation needs, opt-out requests, and concerns about Zenna's use of AI can be sent to:

Rokstars Global HRM Inc.
Email: team@zennahr.com
Website: zennahr.com

18. Reference framework

This statement was informed by the following official guidance. These sources do not constitute a certification or legal endorsement of Zenna.