AI & Automated Decision-Making Notice

1. AI features

• AI-assisted question generation for employers (depends on plan).
• AI-suggested grades for short and long written answers, when the assessment enables it.
• On-device checks for face presence, on-screen gaze and voice comparison.
• Rule-based integrity events and an integrity summary band for reviewers.
• CV parsing and skill-match scores, used for sorting only.

2. How AI is used

When cloud language-model providers are enabled (OpenAI, Google Gemini or DeepSeek), they receive text only: topics, question drafts, written answers and rubrics. Camera and microphone checks run in your browser. We do not send raw video or audio recordings to language-model providers.

3. Evaluation

AI suggests grades for eligible written answers. Coding answers are graded by sandbox tests, and multiple-choice answers by fixed answer keys. Recruiters can override any grade.

4. Scores and recommendations

Scores, integrity bands, match percentages and stage suggestions support the employer's decisions. They are never a binding automated decision by Edgeryt.

5. Human review

A trained person at the hiring company reviews every AI grade and integrity flag before it affects a decision. No candidate is rejected solely by automated means.

Human-only review: before or after the test, you can ask for your written answers to be graded without AI. Use the link on the start screen or email privacy@edgeryt.com. We pass on your request within 2 business days.

6. Proctoring flags

Flags can be raised for: no face, multiple faces, looking away from the screen, other voices, virtual cameras, automation tools or an attached debugger. A flag creates an event for human review. A flag on its own is not proof of cheating.

6A. What we do not do

We do not infer emotions, personality or honesty from your face, voice or gaze. Gaze checks only detect whether you are looking at the screen. We do not train AI on your camera or voice data.

7. Limitations

AI can be wrong, biased or incomplete. Monitoring cannot see a phone outside the camera's view. Grades and bands are not ground truth.

8. Contesting a result

Ask the hiring company, or us, for a human review. You can give your view, and reviewers can regrade.

9. Your rights

• GDPR / UK GDPR: you can obtain human intervention in, give your view on, and contest automated decisions.
• California, from 1 January 2027: businesses using automated decision-making technology for hiring must give a pre-use notice, let you opt out or appeal to a human with authority to overturn the decision, and explain on request how it was used.
• Non-retaliation: using any of these rights will not be held against you.

10. Data used by AI

• Cloud language models: question, answer and rubric text.
• On-device checks: camera and microphone streams, processed locally.

11. Providers

Listed in the Subprocessor List. Only providers enabled on the deployment process data.

12. Fairness and bias audits

We test AI grading and proctoring flags for differences by sex, race/ethnicity and their intersection. We use de-identified data only, and special-category data only where strictly necessary.

New York City (Local Law 144): {{OWNER: EITHER "An independent auditor completed a bias audit on <date>; the summary is at <URL>." OR, until an audit exists, "No independent bias audit has been completed yet. Employers must not use AI grades to substantially assist hiring decisions for New York City roles until one is published."}}

Employers in NYC must give candidates at least 10 business days' notice and explain how to request an alternative process.

13. US state notices

• Illinois: employers must give notice when AI is used in recruitment or hiring (Illinois Human Rights Act 775 ILCS 5/2-102(L), in force since 1 January 2026). If AI analyses a recorded video interview, you get an explanation and are asked for consent first (820 ILCS 42).
• Other states: we update this notice as new laws take effect.

14. EU AI Act

AI used in recruitment is high-risk under Annex III §4. Those obligations apply from 2 December 2027. As provider, Edgeryt is building its risk management, data governance, logging and human-oversight documentation, and gives employers instructions for use. Employers must tell their staff and candidates about the AI they use. Emotion recognition in the workplace is prohibited, and we do not do it.

15. Contact

Platform questions: privacy@edgeryt.com. Complaints about hiring decisions: contact the employer named on your invite.