# AI Transparency Notice.

> How Scrini AI uses AI to source, screen and assess candidates, what it considers, how people stay in charge and the rights candidates have under AI hiring laws.

URL: https://landing.qa.scrini.ai/ai-transparency  
Breadcrumb: Home > AI Transparency Notice

Last updated: October 2, 2026. How our AI agents work in hiring, what they consider, where people stay in charge and the choices candidates have.

## Summary

- Candidates are told when they are speaking with an AI agent.
- Every candidate for a role gets the same questions and the same scorecard, and every score shows the evidence behind it.
- AI agents recommend. The employer's team reviews the evidence, can override any score and makes the hiring decision.
- Customer candidate data is never used to train or fine-tune shared AI models.
- Candidates can ask the employer for an explanation, a human review, an accommodation or an alternative process.

## 1. Why this notice exists

AI now runs much of the first round of hiring on Scrini: sourcing, outreach, phone screens, video interviews, assessments and ranking. Candidates and customers deserve to know how that works. This notice describes our AI features, the safeguards around them and the laws we design for. It supports our customers, who as employers are responsible for how they use AI in their hiring, and it is written for candidates too.

## 2. Our principles

Be open: candidates are told when an AI agent is running a screen or assessment. Be consistent: everyone for a role gets the same core questions and is scored on the same scorecard. Show the evidence: every score links to transcripts, answers or task results. Keep people in charge: recruiters and hiring managers review, override and decide. Protect data: candidate data is used only for that employer's hiring and never to train shared AI models.

## 3. Where AI is used

Each employer chooses which features it turns on.

| Feature | What the AI does | Who decides |
| --- | --- | --- |
| Smart job setup | Turns a role description into must-haves, screening questions and scoring rubrics | The employer reviews and edits before publishing |
| Sourcing and matching (Omni-Source, Neural Match) | Searches the sources the employer chooses and scores how well profiles match the role, with reasons | The employer chooses who to contact or sets rules for outreach |
| Outreach agent | Writes and sends personalized messages, answers questions and follows up by email, SMS, WhatsApp and phone | Candidates can opt out at any time |
| AI phone screening | Runs a structured voice interview, transcribes it and scores answers on the role's scorecard | The employer reviews transcripts and scores |
| AI video interviews (Video Agents) | Runs a two-way video interview with adaptive follow-up questions and produces a structured evaluation | The employer reviews recordings, transcripts and evaluations |
| Assessments and Pixel-Native | Scores tests, code and tasks against a rubric and flags signs of outside help during technical assessments | The employer reviews scores and flags |
| Smart Rank and shortlists | Combines results into a ranked shortlist with the reasons and evidence for each score | The employer reviews, overrides and decides |
| Behavioral HUD | Shows interviewers live notes during interviews: rubric coverage, technical gaps and suggested follow-up questions, plus a summary afterwards | The interviewer decides what, if anything, the notes mean |
| Liveness Verify | Checks that a real person is present and flags possible proxies and deepfakes | Flags go to the employer for review |
| Auto-scheduling | Finds interview times across calendars | Candidates choose a slot |

## 4. What the AI considers

Scores are based on the role's requirements as the employer defines them and on what the candidate says, writes or builds during the process: skills, experience, answers and task results. Scoring criteria must be job-related. Employers must not use protected characteristics such as race, ethnicity, sex, age, disability, religion or sexual orientation as scoring criteria, or proxies for them such as zip codes or photos, as our Acceptable Use Policy explains.

## 5. Voice and video interviews

In AI phone screens and video interviews, the candidate's answers are transcribed and scored against the role's rubric, and the transcript and recording are kept as evidence for each score. Liveness Verify checks that a real person is present and flags possible proxies or deepfakes for the employer's team to review. Where a law requires notice, an explanation or consent before an AI-analyzed video interview or an identity check, such as the Illinois Artificial Intelligence Video Interview Act or the Illinois Biometric Information Privacy Act, the employer is responsible for meeting it, and Scrini provides the information about how the AI works that the employer needs.

## 6. People stay in charge

AI agents recommend and people decide. Scrini does not make final hiring decisions; employers' recruiters review results and decide. The employer's team can review the evidence behind every score and override it, and every override records who made it and when. Employers can set score thresholds that help recruiters prioritize who to review first. Candidates can ask the employer for a human review of any AI-assisted result.

## 7. Explanations

Every score comes with the evidence behind it: the transcript, the answers, the task results and the skills matched or missing. Candidates can ask the employer to explain how an AI-assisted result was reached and which main factors it reflected.

## 8. Fairness testing and bias audits

Structured questions, one scorecard per role, recorded evidence and human review are designed to make screening more consistent than ad hoc calls. We evaluate our models for adverse impact and support customers who must commission independent bias audits (for example under New York City Local Law 144). Where a law requires a bias audit, the employer is responsible for commissioning it and publishing the results.

## 9. Models and training data

Customer candidate data is used only for that customer's hiring and is never used to train or fine-tune shared AI models. Some features use AI models from third-party providers, such as ElevenLabs for AI voice when enabled. Providers that process customer data act as our subprocessors under data processing terms, as described on our Subprocessors page.

## 10. Candidate choices

Candidates are told when an AI is used and can ask the employer for an accommodation, an alternative assessment or a conversation with a person instead of an AI agent, an explanation of a result, a human review of a result, correction of inaccurate data and deletion of their data where the law allows. Candidates can stop outreach at any time.

## 11. Laws we design for

Employers remain responsible for complying with the laws where they hire. This is a summary, not legal advice.

| Law | What it requires | How Scrini helps |
| --- | --- | --- |
| New York City Local Law 144 | For automated employment decision tools: an independent bias audit within one year before use, a public summary of results, notice to NYC candidates at least 10 business days before use and information on how to request an alternative process or accommodation | Recorded evidence, human review and information to support the employer's independent bias audit. The employer is responsible for the audit and for candidate notices |
| Illinois AI Video Interview Act | Before an AI-analyzed video interview: notice, an explanation of how the AI works and what it evaluates, and consent. Limits on sharing videos and deletion within 30 days of a request | Disclosure that the candidate is speaking with an AI, information the employer can use to explain how the AI works and deletion on the employer's instruction. The employer gives notice and obtains consent |
| Illinois Human Rights Act (HB 3773), from January 1, 2026 | Employers must not use AI in a way that discriminates, must notify employees and applicants about AI use and must not use zip codes as a proxy for protected classes | Job-related scoring criteria, human review and disclosure of AI use |
| Colorado AI law | Rules for high-risk AI used in consequential decisions, including employment. The law has been amended and its start date delayed, so employers should check the current requirements and effective date | Notices, explanations and human review |
| California | Civil Rights Council rules on automated decision systems in employment (from October 1, 2025), including keeping records for four years. CCPA rules on automated decision-making technology, including pre-use notice and opt-out or appeal rights, apply from January 1, 2027 | Records of every step, notices and human review |
| EU AI Act | AI used for recruitment and candidate evaluation is high-risk. Emotion recognition in the workplace has been prohibited since February 2, 2025. Duties to tell people they are interacting with an AI and the high-risk obligations apply on the dates set by the Act and any amendments to it | AI disclosure in every interaction, human oversight, logging and documentation that support the employer's own obligations |
| GDPR, UK GDPR and the UK Data (Use and Access) Act 2025 | Safeguards for decisions based solely on automated processing: information, the right to human intervention, to express a view and to contest the decision | Human review and evidence behind every score |
| Quebec Law 25, Brazil LGPD and similar laws | Notice of decisions based exclusively on automated processing and the right to an explanation and review | Explanations and human review through the employer |

## 12. Questions and changes

Candidates should contact the employer first. Anyone can email our privacy team at hello@scrini.ai (subject: Privacy) or support@scrini.ai with questions about our AI. We update this notice when our AI features or the law change, and the date at the top shows the latest version.

## Related policies.

### Privacy Policy

How Scrini collects, uses and protects personal data.

[Read](https://landing.qa.scrini.ai/privacy-policy)

### Candidate Privacy Notice

For people assessed through Scrini by an employer.

[Read](https://landing.qa.scrini.ai/candidate-privacy-notice)

### Cookie Policy

Cookies, consent and how to change your choices.

[Read](https://landing.qa.scrini.ai/cookie-policy)

### Terms of Service

The agreement for using the Scrini platform.

[Read](https://landing.qa.scrini.ai/terms-of-service)

### Acceptable Use Policy

What customers may and may not do with Scrini.

[Read](https://landing.qa.scrini.ai/acceptable-use-policy)

### Data Processing Addendum

Key terms for processing personal data on your behalf.

[Read](https://landing.qa.scrini.ai/dpa)

### Subprocessors

Third parties that process data for Scrini.

[Read](https://landing.qa.scrini.ai/subprocessors)

### Data Security Policy

How we protect data and respond to incidents.

[Read](https://landing.qa.scrini.ai/data-security-policy)

### Refund Policy

Credits, billing disputes and refunds.

[Read](https://landing.qa.scrini.ai/refund-policy)

### Accessibility Statement

Our accessibility commitment and accommodations.

[Read](https://landing.qa.scrini.ai/accessibility)

### Vulnerability Disclosure

How to report a security issue.

[Read](https://landing.qa.scrini.ai/vulnerability-disclosure)