RESPONSIBLE AI STATEMENT

AND ALGORITHMIC TRANSPARENCY DISCLOSURE

PART I: INTRODUCTION AND COMMITMENT

1. Purpose and Scope

This Responsible AI Statement ("Statement") describes how HyreMynd AI Inc. ("HyreMynd," "we," "us," or "our") uses fixed rules, artificial intelligence and automated processing in its hiring platform, including the Workplace Fit Assessment (the "Service"), and the limits of those uses.

It is written with the following laws in mind, where they apply to the actual use, population and jurisdiction. It is not a certification of compliance with any of them:

• (a) Quebec's Act respecting the protection of personal information in the private sector, as amended by Law 25, including its rules on decisions based exclusively on automated processing
• (b) Canada's Personal Information Protection and Electronic Documents Act ("PIPEDA")
• (c) New York City Local Law 144 of 2021 ("NYC LL144")
• (d) Colorado automated decision-making legislation (SB26-189, developer documentation provisions beginning January 1, 2027, where applicable)
• (e) Illinois Artificial Intelligence Video Interview Act ("AIVIA") and Illinois Human Rights Act amendments (HB 3773)
• (f) Title VII of the Civil Rights Act of 1964
• (g) Americans with Disabilities Act ("ADA") and Age Discrimination in Employment Act ("ADEA")
• (h) Other applicable federal, state, provincial, and local laws

2. Our Commitment

HyreMynd builds the Service to be:

• Rule-based where it assesses — Assessment positions and comparisons come from fixed, reproducible rules, not from a model
• Transparent — Clear explanations of system logic, data use, and outputs
• Human-decided — Every output is decision support, and a person makes every employment decision
• Privacy-Protective — Data minimization and strong security controls
• Accountable — The record shows whether a conclusion was a person's own or adopted from an AI draft

PART II: SYSTEM DESCRIPTION AND FUNCTIONALITY

3. Description of the System

The Workplace Fit Assessment does not use AI to score answers. Forced-choice answers are converted into A, B or C working-style positions by fixed majority rules, with ties treated as balanced. Comparisons with a role, a manager and a team, and the outcome categories drawn from them, are computed from those positions by fixed rules. An AI model may then write an explanation of those results; it cannot change them.

AI models are also used to read a résumé or professional profile against a role's posted requirements and propose a reading of each requirement and a qualification category (match, worth a conversation, or doesn't fit); to read public pages found by open-web sourcing and cite the sentence behind each reading, without producing a score; to draft assistant replies and correspondence; and to draft interview notes from an existing transcript.

Interviews can be scheduled on a connected Google or Microsoft calendar. HyreMynd never starts a recording or transcription; where a participant enabled transcription and transcript access was granted, HyreMynd may retrieve the transcript to draft notes for a person to review. Privacy Policy sections 3.7 and 3.8 describe the meeting and transcript flows.

4. Intended Use and Limitations

Use outputs alongside the underlying evidence, lawful selection criteria and accommodations. Never treat a generated reading, comparison or explanation as the sole basis for an employment decision. Generated text can be incomplete or wrong, and rule-based results depend on limited inputs and design choices; none of these outputs is a diagnosis or a scientifically validated measure of personality or ability.

PART III: HUMAN OVERSIGHT AND DECISION-MAKING

5. Human-in-the-Loop Requirements

5.1 Mandatory Human Review

• No automated rejection or advancement without human review
• Human decision-makers must independently evaluate AI outputs
• Reviewers may override or disregard AI recommendations
• Final decisions must always be made by humans

5.2 Substantial Human Involvement

Customer organizations must:

• Assign qualified reviewers
• Train users on AI interpretation
• Document human review processes
• Consider interviews, references, and qualifications
• Retain authority to diverge from AI outputs

5.3 Provenance of Recorded Conclusions

The platform records how each recorded interview conclusion reached a candidate's record. A conclusion recorded with no draft displayed is stamped as human. Where a draft was displayed, the record distinguishes a reviewer who adopted it from a reviewer who overrode it, and that distinction is resolved against the draft the form actually displayed, not against what the reviewer's browser reports.

Agreement with an abstention is not recorded as an unaided human conclusion. Where the model declined to call a dimension and the reviewer answered the same way, the record is stamped ai_concurred rather than human; where the reviewer answered anyway, it is stamped ai_resolved. Neither counts toward the rate at which the model's proposals are accepted, and human therefore means what it says: no draft was in front of the reviewer at all.

6. Right to Human Review and Appeal

Individuals may:

• Request human review
• Appeal adverse decisions
• Correct personal information
• Receive explanations of decision factors
• Contest decisions via appropriate channels

PART IV: TRANSPARENCY AND DISCLOSURE

7. Pre-Assessment Notice Requirements

Before participating, review the notice of the activity and the assessment consent. Account legal acknowledgements and optional provider permissions are separate activities.

8. Explanation of Decision Factors

8.1 Quebec Law 25

Upon request, individuals receive:

• Personal information used
• Principal decision factors
• System parameters
• Correction rights

8.2 Colorado and Connecticut

Where applicable, Colorado SB26-189 replaces the earlier SB24-205 framework and introduces developer documentation requirements beginning January 1, 2027. Connecticut CART provisions phase in from October 1, 2026. Apply the enacted requirements to the actual functions and populations; this Statement is not a compliance certification. Official references: Colorado SB26-189, Connecticut Attorney General notice.

PART V: BIAS PREVENTION AND TESTING

9. Commitment to Non-Discrimination

HyreMynd prohibits discrimination based on protected characteristics, including race, sex, age, disability, religion, gender identity, sexual orientation, veteran status, and others protected by law. Zip codes are not used as proxies.

10. Operational Governance and Assurance Evidence

The Employer determines which employment-tool rules apply to its actual use, population and jurisdiction, and owns required candidate notices, alternative procedures and bias-audit procurement and publication. HyreMynd provides available technical and processing documentation to support those obligations. Independent audits, published summaries, demographic monitoring, organizational training, impact assessments and certifications require operational evidence. This Statement does not certify any of those activities as completed, and does not present an engineering test as a bias audit. Request the applicable dated evidence and contractual commitments from legal@hyremynd.com.

11. Ongoing Monitoring and Improvement

Material changes to required legal documents request a new acknowledgement of the exact version and content. Technical changes and assurance evidence are reviewed separately under Section 10.

PART VI: DATA GOVERNANCE

12. Data Collection and Use

The Service uses the data needed for the selected feature, including professional networking imports, employer-directed open-web sourcing, résumés, answers and authorized interview transcripts. Assessment derivation uses only the actual forced-choice answers; social-media browsing is not an input to it. Employers must not supply protected characteristics or unlawful proxies as selection criteria. The presence of professional sourcing does not authorize unrelated social profiling.

13. Data Retention and Deletion

Retention follows the category schedule in Privacy Policy section 9, and erasure follows the staged process in Privacy Policy section 10.3.

PART VII: SECURITY AND RISK MANAGEMENT

14. Security Measures

Technical controls include encryption, tenant isolation, access controls and operation logging. Penetration testing, employee training, incident programs and independent assurance must be evidenced separately under the applicable service agreement; passing code tests does not prove those activities.

PART VIII: REGULATORY COMPLIANCE

15. Algorithmic Discrimination Reporting

Report suspected algorithmic discrimination to legal@hyremynd.com. HyreMynd reviews each report, corrects a defect it confirms in the Service, and informs the affected Employer. Notices to authorities or to individuals follow the law that applies to the controller concerned.

16. Recordkeeping

Records are kept for the periods in Privacy Policy section 9. There is no blanket multi-year minimum for assessments, transcripts or logs. Required legal preservation uses a recorded legal hold with a reason, a review date and an end date; minimized consent, mail and request proof each have their own purpose and period.

PART IX: CUSTOMER AND CANDIDATE RIGHTS

17. Candidate Rights Summary

Includes rights to notice, explanation, human review, appeal, correction, accommodations, deletion, and withdrawal of consent.

18. How to Exercise Rights

Requests may be submitted via:

• The Employer running the hiring process
• privacy@hyremynd.com
• The contact details in the pre-assessment notice
• HyreMynd's Privacy Officer

PART X: DISCLAIMERS

19. System Limitations

• No HR, legal, medical, or psychological advice
• No guaranteed outcomes
• Not compliance certification
• Not error-free or bias-free
• Must be one factor among many
• Transcripts are produced by the meeting platform, not by HyreMynd; their accuracy varies with audio quality, accents, overlapping speech, and language, and a draft derived from an inaccurate transcript inherits that inaccuracy
• Transcript ingestion works only on the meeting platforms HyreMynd integrates with, and only where the account holder has granted transcript access
• Most interviews produce no transcript at all — neither platform transcribes a meeting nobody enabled transcription for, so the absence of a transcript-derived draft is the ordinary outcome and says nothing about the candidate

PART XI: UPDATES AND CONTACT INFORMATION

20. Updates

This Statement is versioned. The effective date and version shown with it identify the current text, and earlier versions remain available in the legal document archive.

21. Contact Information

HYREMYND AI INC.

Attention: Privacy Officer / AI Governance

559 Sammon Ave, East York, Ontario, M4C 2E1, Canada

📧 privacy@hyremynd.com

22. Regulatory Authority Contacts

• Quebec CAI — www.cai.gouv.qc.ca
• OPC Canada — www.priv.gc.ca
• NYC DCWP — www.nyc.gov/dca
• Colorado AG — coag.gov
• Illinois DHR — www2.illinois.gov/dhr
• EEOC — www.eeoc.gov

— END OF RESPONSIBLE AI STATEMENT —