Security and Trust

• Team-scoped access. Data is isolated by team. Recruiters and team members can only view, edit, or manage candidates and positions belonging to their own organisation. Cross-tenant access is blocked at the database level using row-level security.
• Role-based permissions. Each team member is assigned a role (admin or user) that determines what actions they can perform, such as inviting members, managing billing, or overriding AI decisions.
• Server-side verification. Every data access request is verified server-side. The caller's team membership is checked against the position or application being accessed before any data is returned or modified.

Hiring teams can configure automatic deletion of application data after a set period. The default is 6 months, but this is turned off until a team enables it. Teams can choose any period from 1 to 60 months. When auto-delete is enabled, application data (including recordings, transcripts, AI analyses, resumes, and answers) is removed after the configured period. A bare record (candidate name, stage history, and scorecard decisions) is retained for hiring-records purposes.

Candidates can request deletion of their data at any time. Every deletion is recorded in an audit log. For more details, see our Privacy Policy.

Before any resume text is sent to an AI model, we automatically remove personal contact details. This includes email addresses, phone numbers, URLs, street addresses, and the candidate's own name. Only the redacted text is used in AI prompts.

• Disclosure before applying. Candidates are informed when AI will be used in the screening process before they submit an application. They must acknowledge this disclosure to proceed.
• Right to a human-only review. Candidates can opt out of AI scoring entirely. When they do, no AI fit score is generated. Their rounds are evaluated by a human recruiter using a manual scorecard.

During AI-led interview rounds, the platform records objective integrity signals: tab switches, time away from the page, and paste events. Only counts, durations, and timestamps are stored. No clipboard contents, keystrokes, or screen recordings are captured. These signals are shown to the recruiter as context only. They never affect the candidate's score, grade, stage, or any automated decision.

• Demographic data. Candidates can optionally provide demographic information (sex and race/ethnicity) for bias-audit purposes. This data is visible only to team admins.
• Impact ratio analysis. Teams can calculate selection rates and impact ratios to identify potential adverse impact under the four-fifths (80%) rule.
• Audit reports. Bias-audit reports capture the methodology, demographic breakdowns, and findings. These are visible to team admins for ongoing monitoring.

Audit results are informational and do not automatically alter AI scoring.

• Advisory only. AI-generated scores are advisory only. A human recruiter reviews all results and makes the final hiring decision.
• Overrides. Recruiters can override any AI recommendation with a documented reason. The override and the reason are recorded in the audit log.
• Three interview modes. Recruiters can configure each round as AI-led, a live human video interview, or a hands-on exercise. In human and hands-on rounds, the recruiter's own scorecard is the hiring decision.

For more information, see our AI Transparency page.

Users can sign in with an email address and password, or with Google. Passwords are managed by the platform's authentication system. We do not store passwords in plain text.

The platform records an audit log of key actions, including AI screening events, human overrides, stage changes, application views, team member changes, and data erasure actions. The audit log is visible to team admins.

If you believe you have found a security vulnerability or have a security concern, please report it to us at support@robrics.ai. Please do not share the details publicly until we have had a chance to investigate.

See also our Privacy Policy, Terms of Service, and Sub-processors.