Skip to content
Vocaid

Subscribe to our newsletter

Weekly digest of new blog posts. Unsubscribe anytime.

Vocaid

Turn interviews into measurable progress.

Product

  • Interview Practice
  • Refer & earn credits

Developers

  • API Reference

Company

  • About

Legal

  • Privacy Policy
  • Terms of Use
  • AI Transparency

© 2026 Vocaid. All rights reserved.

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

Vocaid — Rua Pais Leme, 215, Conj 1713, Pinheiros, São Paulo/SP
CNPJ 65.669.479/0001-07

    AI Transparency & Candidate Rights

    Last updated: July 6, 2026

    Table of Contents

    • Overview
    • AI Systems Inventory
    • Scoring Methodology
    • Bias Prevention
    • Emotion Recognition Prohibition
    • Regulatory Compliance
    • AI Governance Framework
    • Candidate AI Rights
    • EU AI Act Compliance (Annex IV)
    • NYC Local Law 144 Compliance
    • Third-Party AI Services
    • Payment Processing
    • Contact Us

    1. Overview

    Vocaid is committed to transparency about how artificial intelligence is used in our platform. This page provides a comprehensive overview of our AI systems, how they generate outputs, their known limitations, and your rights as a user.

    We believe that every person who interacts with AI-powered hiring tools has the right to understand how those tools work and how they may affect decisions about their career.

    2. AI Systems Inventory

    Vocaid uses the following AI systems in its platform. Each system is described with its purpose, the data it processes, and the safeguards in place.

    By default, your interview recordings and transcripts are deleted on our standard schedule. You may optionally opt in to extended retention so this data can be used to improve interview analytics and AI quality; this is off by default and can be withdrawn anytime in your account settings.

    Two systems below — Behavioral Analysis and Liveness Detection — are currently disabled on Vocaid. While they are disabled, no camera video leaves your device for these features and no facial or biometric processing takes place. We keep their descriptions published so you can see exactly what they do before we ever re-enable them; if we re-enable either one, we will update this page and ask for the required consents first.

    Vocaid also suggests and ranks job listings by comparing role requirements with your profile, resume, and preferences (using semantic text-similarity, not an AI judgment of you). These suggestions are advisory, are shown only to you, and share nothing with employers or job boards; see the Privacy Policy section "Job Listings and Matching".

    Voice Interview AI

    Conducts real-time voice-based interviews using natural language processing

    Data Processed

    • •Audio input from microphone
    • •Job description and resume content
    • •Conversation context from current session

    Outputs

    • •Natural language interview questions
    • •Follow-up questions based on responses
    • •Real-time speech-to-text transcription

    Known Limitations

    • •May misunderstand heavy accents or dialects
    • •Transcription accuracy varies with audio quality
    • •Cannot assess physical demonstrations or hands-on skills
    • •Response quality depends on internet connection stability

    Human Oversight

    All AI-generated scores are presented as recommendations. B2B hiring managers retain full decision authority. Candidates may request human review.

    Interview Scoring & Feedback

    After each interview, generates your feedback report — a summary, dimension scores, strengths, improvement areas, and recommendations — from the transcript of what you said.

    Data Processed

    • •The interview transcript (your words) and the question-and-answer history
    • •Your resume and the job description, used as context

    Outputs

    • •An interview summary
    • •Five dimension scores from 0 to 100 — communication, problem-solving, technical depth, leadership, clarity — each with its reasoning
    • •An overall score from 0 to 100 that combines the dimensions, weighted for the seniority of the target role
    • •Strengths, improvement areas, and recommendations

    Known Limitations

    • •Scores only the content of the transcript — never your tone of voice, accent, face, appearance, or emotions
    • •A very low score on a core dimension caps the overall score, so one weak area stays visible instead of being averaged away
    • •If an interview cannot be scored, it is shown as unscored — a score is never invented
    • •Scores are relative coaching assessments and can vary between sessions

    Human Oversight

    Scores are advisory. For personal practice they are coaching feedback only. Wherever an employer uses interview results, the hiring decision is always made by a human, and you can request human review of any score.

    Behavioral Analysis

    Analyzes non-verbal communication signals during video interviews for coaching feedback

    Data Processed

    • •Video frames (sampled, not continuously stored)
    • •Face detection coordinates (not facial recognition)
    • •Gaze direction estimates
    • •Head position and movement patterns

    Outputs

    • •Eye contact consistency percentage
    • •Posture and engagement indicators
    • •Response timing metrics
    • •Gaze tracking heatmap data

    Known Limitations

    • •Currently disabled: this feature is switched off for all practice interviews — no video frames are captured or analyzed while it is off
    • •Accuracy depends on camera quality and lighting
    • •Cannot assess emotional state (emotion recognition is prohibited)
    • •May be less accurate for certain face shapes or skin tones
    • •Requires frontal camera position for reliable results

    Human Oversight

    Behavioral signals are supplementary coaching data only. They are never the sole basis for scoring. Users can opt out entirely and use audio-only mode.

    Fit & Readiness (resume check)

    An advisory self-assessment of your resume against a target role or job description. There is no universal "ATS score" — this is a readiness check for you, not a prediction of any employer's screening or hiring decision.

    Data Processed

    • •Your resume text
    • •The job description you provide (optional)
    • •Open skills taxonomies (O*NET and ESCO) used to recognize and normalize skills

    Outputs

    • •An overall readiness score from 0 to 100, combining keyword coverage, format, experience, and education
    • •Keyword coverage against the job description, computed by direct skill overlap (not by AI judgment)
    • •A semantic-fit indicator comparing your resume and the job description (informational)
    • •A parse-fidelity check of how reliably your resume's structure can be read
    • •Keywords found and missing, plus 3–6 written suggestions
    • •A standing disclaimer that this is an advisory check, shown wherever the score appears

    Known Limitations

    • •A heuristic, not a measurement: we hold no data linking resume scores to hiring outcomes, so we make no accuracy claims
    • •The headline number comes from deterministic matching and semantic-similarity layers; the AI writes the explanations and suggestions — it does not produce the score
    • •Cannot verify claims made in resumes, and may undervalue non-traditional career paths
    • •Skill matching may miss equivalent terminology

    Human Oversight

    Candidate-facing only — never used to score you for an employer. Advisory: it does not determine interview access or any hiring outcome.

    Resume Writing Assistance

    Optional AI writing help inside the resume builder: drafts bullet points from your role details, gives feedback on sections you write, suggests edits to tailor your resume to a job description, and structures the content of an uploaded resume into your profile.

    Data Processed

    • •The resume content you are editing, including text you type in the builder
    • •A job description you provide for tailoring suggestions (optional)
    • •The text of a resume file you upload for autofill

    Outputs

    • •Draft bullet points and suggested wording you can accept, edit, or discard
    • •Written feedback and improvement suggestions on your resume sections
    • •Suggested edits to align your resume with a specific job description

    Known Limitations

    • •Suggestions are drafts — review them for accuracy before using them; never present AI-drafted claims you cannot back up
    • •It cannot verify your work history and may produce generic or imprecise wording
    • •Quality depends on the detail you provide; sparse inputs produce sparse suggestions
    • •Currently powered by Azure OpenAI; if we change the model provider, this page and our privacy policy will be updated before the change takes effect

    Human Oversight

    Nothing is added to your resume without your action — you choose whether to accept each suggestion. This assistance is candidate-facing only and plays no role in any employer's screening or hiring decision.

    Liveness Detection

    Verifies that a real person is present during the interview session to prevent fraud

    Data Processed

    • •Video frames during verification
    • •Face detection coordinates
    • •Head movement tracking

    Outputs

    • •Pass/fail liveness verification result
    • •Confidence score (internal use only)

    Known Limitations

    • •Currently disabled: this check is switched off — no face detection or liveness processing runs while it is off
    • •May fail in poor lighting conditions
    • •Requires camera access (can be skipped with fallback)
    • •May be less reliable with certain accessories (masks, heavy makeup)

    Human Oversight

    Liveness verification is a binary pass/fail check. Failed verification can be retried or skipped. It does not affect interview scoring.

    AI Coach Voice (synthetic speech)

    Optionally reads the AI Coach's onboarding tips and your feedback summary aloud using an AI-generated synthetic voice (ElevenLabs text-to-speech). Speech output only; this feature is opt-in and off by default.

    Data Processed

    • •Text of coaching tips and onboarding prompts (fixed, pre-written lines)
    • •A short feedback summary derived from your own interview scores and feedback, converted to text and sent for speech synthesis
    • •No microphone audio or recording of your voice is sent for this feature — it is text-to-speech (output only)

    Outputs

    • •Synthetic spoken audio of the coaching tip or feedback summary

    Known Limitations

    • •Synthetic, AI-generated voice — not a recording of a real person (EU AI Act Article 50 disclosure)
    • •May mispronounce names, acronyms, or uncommon words
    • •Opt-in and off by default; the same content is always available as on-screen text
    • •The spoken content reflects AI-generated coaching text and is informational, not professional career advice

    Human Oversight

    The synthetic voice only reads back coaching text the AI Coach already produced; it makes no decisions. You choose whether to play audio, and the equivalent text is always shown on screen.

    Onboarding Assistant (Vocaid Coach / HR Assistant)

    Conversational AI companion that guides new users through in-app onboarding tours, highlighting interface elements and answering questions you type to it

    Data Processed

    • •Messages you type to the assistant during a tour
    • •Your quick-reply selections
    • •Which onboarding tour is active and its step progress (server-side, not typed by you)

    Outputs

    • •Short guidance messages in the chat bubble
    • •Interface actions such as highlighting a button or marking a step complete

    Known Limitations

    • •Scoped to onboarding topics; off-topic answers are brief and redirect back to the tour
    • •May occasionally misunderstand a typed question
    • •Makes no decisions about you, your account, or your interview results

    Human Oversight

    The assistant only guides navigation and never makes decisions about you. It is clearly labeled as AI in the chat header, so you always know you are interacting with an AI and not a human. You can dismiss or defer the tour at any time.

    Performance Coach Chat ("Ask anything about your performance")

    Conversational AI assistant that answers questions you type about your own interview performance, feedback, and preparation. It uses your interview history as context so its answers are specific to you.

    Data Processed

    • •Messages you type to the assistant
    • •Your interview transcripts, scores, and feedback, used as context for answers
    • •Your resume text and profile details, used as context

    Outputs

    • •Conversational coaching answers about your performance and preparation
    • •Suggested focus areas for your next practice sessions

    Known Limitations

    • •Informational coaching, not professional career advice — it may occasionally be inaccurate or generic
    • •Sees only your own data; it has no access to other candidates or to employers
    • •Conversations are stored with your account and automatically deleted after 90 days (see our Privacy Policy); they are not used to train the underlying AI models
    • •Currently powered by Azure OpenAI; if we change the model provider, this page and our privacy policy will be updated before the change takes effect

    Human Oversight

    The assistant is clearly labeled as AI. It makes no decisions about you, your account, your scores, or any hiring outcome — it only explains and coaches. Using it, or not using it, never affects your scores or feedback.

    3. Scoring Methodology

    Vocaid's interview scoring evaluates candidates across multiple dimensions using AI analysis of their spoken responses. Scores reflect the AI's assessment of demonstrated competencies, not the person's inherent abilities.

    Scoring Dimensions

    • Communication: how effectively you articulate ideas, listen, and engage in dialogue
    • Problem-Solving: how methodically you approach problems, think critically, and propose solutions
    • Technical Depth: the depth and accuracy of the domain and technical knowledge you demonstrate
    • Leadership: ownership, decision-making, and capacity to influence, as demonstrated in your answers
    • Clarity: clarity of thought, structured responses, and concise expression

    Scores range from 0 to 100 and represent relative assessments. They should be interpreted as coaching feedback, not absolute measurements. Scores may vary across sessions due to question variation, AI model updates, and response differences. The overall score combines the five dimensions, weighted for the seniority of the target role. If any core dimension (communication, problem-solving, or technical depth) scores very low, the overall score is capped so a weak area stays visible. If an interview cannot be scored, we show it as unscored rather than inventing a number.

    4. Bias Prevention

    We are committed to ensuring our AI systems do not discriminate against any individual or group based on protected characteristics including race, gender, age, disability, national origin, or other legally protected categories.

    Our Bias Prevention Measures

    • Regular testing of AI outputs across demographic groups to identify disparate impact
    • Prohibition of emotion recognition in workplace assessment contexts (EU AI Act compliance)
    • Scoring based on content and communication quality, not voice characteristics, accent, or appearance
    • Continuous monitoring of score distributions for statistical anomalies
    • Planned independent bias audit by a qualified third-party auditor (per NYC Local Law 144 requirements)

    If you believe you have experienced bias in AI-generated scores or feedback, please report it to support@vocaid.ai. We investigate all bias reports and take corrective action where warranted.

    5. Emotion Recognition Prohibition

    Vocaid does NOT use emotion recognition, emotion inference, or affective computing in any of its AI systems. This prohibition is a core design principle, not merely a compliance measure.

    The EU AI Act (Article 5(1)(f)), effective February 2, 2025, explicitly prohibits the use of AI systems that infer emotions in the workplace and educational institutions, except for medical or safety purposes. Vocaid's behavioral analysis features produce only factual, observable outputs:

    • Eye contact consistency percentage (measured by gaze direction relative to camera)
    • Posture indicators (head position, shoulder alignment detected from video frames)
    • Response timing metrics (duration between question and answer)
    • Gaze tracking heatmap data (direction of visual attention)

    These outputs describe observable physical behaviors and do NOT infer internal emotional states such as confidence, nervousness, stress, enthusiasm, or any other affective characteristic. Our behavioral analysis is fundamentally different from emotion recognition because it measures what a person does, not what they feel.

    Strict Prohibition

    Vocaid's engineering standards prohibit any model output label that implies emotional inference. All behavioral analysis outputs are reviewed to ensure they describe only factual observations. Violation of the EU AI Act emotion recognition prohibition carries penalties of up to EUR 35 million or 7% of global annual turnover.

    The optional AI Coach synthetic voice (ElevenLabs text-to-speech) is output only: it reads coaching text aloud and never analyzes your voice. Vocaid does not infer your emotions from your voice, and no microphone audio is processed by this feature.

    6. Regulatory Compliance

    Vocaid operates in compliance with AI regulatory frameworks across multiple jurisdictions. Below are the specific obligations and our compliance measures for each applicable regulation.

    EU AI Act Compliance

    Vocaid is classified as a high-risk AI system under Annex III, Category 4 (Employment, Workers Management and Access to Self-Employment) of the EU AI Act (Regulation (EU) 2024/1689). As a provider of AI-powered hiring assessment tools, we are subject to the requirements of Chapter 3, Section 2.

    • Risk Management System (Art. 9): We maintain a documented risk management system that identifies, evaluates, and mitigates risks throughout the AI system lifecycle
    • Data Governance (Art. 10): Training, validation, and testing data sets are subject to data governance and management practices ensuring quality, representativeness, and bias detection
    • Technical Documentation (Art. 11): We maintain comprehensive technical documentation describing our AI systems, their capabilities, limitations, and the methodologies used
    • Record-Keeping (Art. 12): Automatic logging of AI system operations enables traceability of system behavior
    • Transparency (Art. 13): This AI Transparency page fulfills our transparency obligations by providing clear information about system capabilities and limitations
    • Human Oversight (Art. 14): All AI-generated scores are advisory. Human decision-makers (hiring managers) retain authority over employment decisions. Override capabilities are built into B2B workflows.
    • Accuracy, Robustness, and Cybersecurity (Art. 15): We implement measures to ensure appropriate levels of accuracy, robustness, and cybersecurity
    • Fundamental Rights Impact Assessment (Art. 27): We commit to conducting fundamental rights impact assessments before deploying in EU markets

    Conformity assessment and EU database registration (Art. 60) are in progress. Under the Digital Omnibus amending regulation adopted by the European Parliament and Council in June 2026, the application date for high-risk AI obligations (Annex III, including employment) is postponed from August 2, 2026 to December 2, 2027, subject to publication in the Official Journal.

    Emotion Recognition

    The prohibition on emotion recognition in workplace AI systems (Art. 5(1)(f)) has been in effect since February 2, 2025. Vocaid has never used emotion recognition and confirms full compliance with this prohibition.

    Colorado AI Act (SB 26-189)

    The Colorado AI Act (SB 26-189), effective January 1, 2027, establishes obligations for developers and deployers of automated decision-making technology used in consequential decisions. It repealed and replaced the 2024 law (SB 24-205) before that law took effect. As a developer of AI hiring assessment tools, Vocaid acknowledges the following obligations:

    • Developer Disclosure: We provide deployers (employers using our platform) with documentation describing the intended uses, known limitations, and risk-mitigation measures of our AI systems
    • Transparency Support: We provide the information deployers need to give candidates pre-use notice and to explain adverse decisions
    • Consumer Notification: Candidates are notified when automated decision-making technology makes or substantially influences a consequential decision about them
    • Statement of Compliance: We make available a statement summarizing the automated decision-making systems we develop and how we support fair, transparent use
    • Attorney General Cooperation: We will cooperate with the Colorado Attorney General (the sole enforcer of the Act) and provide documentation upon request

    Vocaid's deployers (employers) have independent obligations under the Colorado AI Act, including providing pre-use notice, disclosing adverse decisions, offering a right to human review, and retaining records. We support deployers in meeting their obligations through our platform documentation and transparency tools.

    Texas Responsible AI Governance Act

    The Texas Responsible AI Governance Act, effective January 1, 2026, requires companies deploying AI systems in Texas to maintain AI governance frameworks. Vocaid complies with the following requirements:

    • AI Governance Framework: We maintain internal documentation of our AI governance practices, available to the Texas Attorney General upon request
    • Prohibited Uses: Our AI systems do not encourage self-harm, criminal activity, or unlawful discrimination
    • Transparency: We provide clear disclosure of AI use in our hiring assessment tools
    • Biometric Data: We clarify when biometric data is collected for AI processing and obtain appropriate consent under Texas law

    NYC Local Law 144 (AEDT)

    Vocaid's AI interview and scoring platform constitutes an Automated Employment Decision Tool (AEDT) under NYC Local Law 144. We comply with the following requirements:

    • Annual Bias Audit: We commit to conducting an independent bias audit by a qualified third-party auditor, analyzing selection rates by sex, race/ethnicity, and intersectional categories
    • Public Summary: The bias audit summary will be posted publicly on our website within 6 months of each audit's completion
    • Candidate Notification: We provide candidates with at least 10 business days' notice before an AEDT is used in an employment decision, as implemented in our candidate notification system
    • Data Disclosure: Candidates can request information about the data collected, the source of such data, and the employer's data retention policy

    Brazil LGPD Article 20

    Under Article 20 of Brazil's Lei Geral de Proteção de Dados (LGPD), data subjects have the right to request human review of decisions made solely through automated processing that affect their interests, including profiling. Vocaid ensures:

    • Candidates assessed by our AI systems can request human review of automated decisions through our Privacy Center or by contacting our Encarregado (DPO)
    • We provide transparency about the criteria and procedures used in automated decision-making
    • All AI scores are presented as advisory recommendations, never as binding decisions

    Mexico LFPDPPP (2025 Reform)

    The 2025 reform to Mexico's Federal Law on Protection of Personal Data Held by Private Parties introduces new obligations for automated decision-making. Vocaid complies with:

    • Automated Decision-Making Transparency: We disclose when automated decision-making technology is used in the assessment of candidates
    • Right to Human Intervention: Data subjects can request human intervention in automated decisions that produce legal effects or similarly significantly affect them
    • Impact Assessment: We commit to conducting impact assessments for automated processing of personal data
    • Enhanced Consent: We obtain specific, informed consent for the processing of personal data through automated decision-making

    Colombia SIC Circular 001/2025

    The Colombian Superintendencia de Industria y Comercio (SIC) issued Circular 001/2025 addressing biometric data processing requirements. For Vocaid's operations in Colombia:

    • Biometric Proportionality: We ensure that any biometric data collection is proportional to the stated purpose and does not exceed what is strictly necessary
    • Purpose Limitation: Biometric data is collected solely for interview verification and communication coaching feedback, not for surveillance or ongoing monitoring
    • SIC Compliance: We commit to registering our databases with the RNBD (Registro Nacional de Bases de Datos) as required by Colombian law

    7. AI Governance Framework

    Vocaid maintains a comprehensive AI governance framework that addresses the responsible development, deployment, and monitoring of our AI systems across all jurisdictions where we operate.

    Our Governance Practices

    • Risk Management: Systematic identification, assessment, and mitigation of risks associated with our AI systems, including risks of bias, discrimination, and inaccuracy
    • Bias Testing and Monitoring: Regular testing of AI outputs across demographic groups to identify and address disparate impact or unfair bias patterns
    • Continuous Monitoring: Ongoing monitoring of deployed AI systems for performance degradation, drift, and emergent bias
    • Model Versioning: Version control and documentation of all AI model updates, with impact assessment before deployment of significant changes
    • Incident Reporting: Internal processes for reporting, investigating, and remediating AI-related incidents, including bias complaints, accuracy issues, and system failures
    • Stakeholder Engagement: Regular engagement with affected stakeholders including candidates, employers, regulators, and civil society organizations

    Our AI governance framework is designed to evolve with the regulatory landscape. We regularly review and update our practices to align with new requirements under the EU AI Act, Colorado AI Act, Texas Responsible AI Governance Act, and other emerging AI regulations.

    8. Candidate AI Rights

    Vocaid recognizes the following rights for all individuals who interact with our AI systems:

    1

    Right to Know

    You have the right to know that AI is being used in your interview process and how it affects your assessment.

    2

    Right to Consent

    You have the right to provide informed consent before biometric data (voice patterns, facial geometry) is processed.

    3

    Right to Opt Out

    You have the right to opt out of video recording and behavioral analysis and use audio-only mode without penalty.

    4

    Right to Human Review

    You have the right to request human review of any AI-generated score or assessment.

    5

    Right to Appeal

    You have the right to contest AI-generated assessments and provide additional context.

    6

    Right to Explanation

    You have the right to receive a meaningful explanation of how your scores were generated.

    7

    Right to Data Access

    You have the right to access all data collected about you and all AI-generated outputs.

    8

    Right to Deletion

    You have the right to request permanent deletion of your interview data, recordings, and AI outputs.

    9

    Right to Non-Discrimination

    You have the right to fair treatment regardless of your race, gender, age, disability, accent, or other personal characteristics.

    10

    Right to Accommodations

    You have the right to request reasonable accommodations for disabilities that may affect AI assessment accuracy.

    6. EU AI Act Compliance (Annex IV)

    In accordance with the European Union AI Act (Regulation 2024/1689), Vocaid provides the following technical documentation for our AI systems classified under the Act. Our systems are designed to comply with transparency obligations for AI systems that interact with natural persons.

    System Information & Intended Purpose

    • Provider: Vocaid, Inc.
    • System Name: Vocaid AI Interview Platform
    • Version: Production (continuously updated)
    • Intended Purpose: AI-powered interview practice, resume scoring, and behavioral communication coaching
    • Intended Users: Job seekers (B2C) and hiring organizations (B2B)
    • EU Representative: Contact support@vocaid.ai for EU representative details
    • Risk Classification: Limited Risk (transparency obligations apply under Article 50)

    Training Data & Model Information

    • Vocaid uses third-party foundation models — Azure OpenAI (GPT-4o and GPT-4o-mini) for scoring, feedback, resume analysis, and coaching chat, and Anthropic Claude models (via OpenRouter) for the interview conversation and assistants — we do not train custom models on user data
    • Resume Fit & Readiness scoring combines deterministic skill matching, semantic text-similarity comparisons (embeddings), and an AI model that writes explanations and suggestions; the AI does not produce the headline score
    • Behavioral analysis uses computer vision algorithms for face detection and gaze estimation, not facial recognition
    • No biometric identification or categorization is performed
    • Emotion recognition is explicitly prohibited in our systems (EU AI Act Article 5(1)(f) compliance)

    Risk Management & Mitigation

    Vocaid implements the following measures to mitigate risks associated with AI-assisted assessment:

    • Human oversight: All AI-generated scores are advisory — hiring decisions are always made by humans
    • Bias monitoring: Regular statistical analysis of score distributions across demographic groups
    • Transparency: Full disclosure of AI involvement before any assessment begins
    • Data minimization: Only data necessary for the specific assessment purpose is collected
    • User control: Users can opt out of video/behavioral analysis at any time without penalty
    • Logging: All AI decisions are logged for auditability per Article 12 requirements
    • Incident reporting: Processes in place for reporting serious incidents to relevant authorities

    For the complete technical documentation file or to submit questions regarding EU AI Act compliance, contact support@vocaid.ai. We are committed to full compliance with the Act's requirements as they enter into force.

    7. NYC Local Law 144 Compliance

    New York City Local Law 144 (2021) requires employers and employment agencies that use automated employment decision tools (AEDTs) to conduct annual independent bias audits and provide notice to candidates. Vocaid is committed to compliance with LL144 requirements.

    Applicability & Scope

    • Vocaid's AI interview scoring system qualifies as an AEDT when used by employers to substantially assist or replace discretionary decision-making in hiring
    • LL144 applies to candidates for employment opportunities in New York City
    • For B2C users (personal practice), the tool is advisory only and does not constitute an AEDT
    • For B2B employers using Vocaid to screen candidates, LL144 requirements apply

    Independent Bias Audit

    Vocaid will commission an independent bias audit before deploying our AI scoring for B2B hiring decisions in NYC. The audit will evaluate:

    • Impact ratio: Scoring rates for each race/ethnicity category compared to the most selected category
    • Impact ratio: Scoring rates for each sex category compared to the most selected category
    • Median scores: Comparison of median AI-generated scores across demographic categories
    • Selection rates: Analysis of pass/fail rates across protected categories at various score thresholds
    • Intersectional analysis: Combined race/ethnicity and sex impact ratios where data is sufficient

    Candidate Notice

    In compliance with LL144, Vocaid provides the following notice to candidates evaluated for NYC employment opportunities: An automated employment decision tool (AEDT) will be used in connection with the assessment of your job application. The AEDT evaluates your interview responses for communication clarity, technical knowledge, problem-solving ability, and relevance. You may request an alternative selection process or accommodation by contacting the employer directly or by emailing support@vocaid.ai.

    The bias audit summary will be published on this page once completed. Vocaid retains bias audit results for at least four years as required by law. For questions about LL144 compliance, contact support@vocaid.ai.

    8. Third-Party AI Services

    Vocaid uses the following third-party AI services. We maintain Data Processing Agreements with each provider to ensure your data is handled responsibly.

    • Azure OpenAI Service (Microsoft) — Powers interview conversation, question generation, and performance scoring↗
    • Deepgram — Real-time voice interview agent (speech-to-text, conversation flow, speech synthesis)↗
    • ElevenLabs — Real-time voice interview agent across all interview languages — speech-to-text, speech synthesis, and conversation relay↗
    • Azure Face API (Microsoft) — Face detection and liveness verification (not facial recognition)↗
    • Anthropic (PBC), via OpenRouter — Powers the interview conversation (question flow and follow-ups, routed via OpenRouter), and — via Anthropic's direct API — the in-app AI assistant/tour coach and content generation. Interview scoring, feedback, and the performance-coach chat run on Azure OpenAI and are routed to Anthropic only where our data processing agreement with Anthropic covers that data.↗
    • OpenRouter Inc. — Routing layer that relays interview-conversation and content-generation prompts to Anthropic Claude models; processes prompt content in transit. It does not receive your resume, scores, or feedback.↗

    9. Payment Processing

    Card payments are handled by our PCI-compliant payment processor. Vocaid never stores raw card numbers. If you choose to connect a cross-border payout provider (Deel or Wise), we store an OAuth token to act on your behalf at your request — we never see your provider login credentials.

    • Stripe — Payment processing (PCI DSS Level 1) — card details are tokenized by Stripe and never reach Vocaid's servers↗
    • Deel — Cross-border candidate payout / EOR (United States) — if you connect your Deel account, we store an OAuth token to sync your contract and payment status at your request (not yet live)↗
    • Wise — Cross-border candidate payouts in your local currency (UK/EU) — if you connect your Wise account, we store an OAuth token to initiate payouts at your request (not yet live)↗

    10. Contact Us

    For questions about our AI practices, to exercise your rights, or to report a concern:

    Email: support@vocaid.ai

    We take all AI-related inquiries seriously and aim to respond within 15 business days.