EU AI ACT
Annex IV public summary
This page publishes a customer-facing summary of Intrvio's AI interview system documentation.
Intended purpose
GAIA supports structured interview collection, transcription, scoring assistance, and reviewer summaries. Employers remain responsible for final hiring decisions.
System architecture overview
Candidate setup, recording, voice interaction, transcript processing, AI analysis, reviewer console, audit logging, and retention jobs are separated by service boundary.
Training data summary
Intrvio does not train foundation models on customer interview content. Evaluation prompts and rubrics are configured per employer template.
Performance metrics
Customers can review completion rate, recording integrity, evaluator version, confidence signals, human review overrides, and bias-audit exports where available.
Risk management overview
Key risks include unfair screening, accessibility barriers, biometric sensitivity, model drift, hallucinated summaries, and overreliance by reviewers.
Human oversight design
GAIA output is advisory. Employer reviewers see transcript evidence and can override or disregard AI recommendations.
Technical robustness
Controls include permission checks, recording reconciliation, audit trails, role-based access, rate limits, and retention workflows.
Monitoring
Operational monitoring covers webhook delivery, recording status, analysis runs, retention jobs, abuse signals, and customer support escalations.
Change log
Material AI system changes are versioned through evaluator versions, template snapshots, release notes, and Trust Center updates.