AI Usage, Output & Transparency Policy

1. Purpose

This policy is the single public source for how TruTalent uses AI, what AI outputs mean, permitted reliance, transparency duties, candidate safeguards, and customer responsibilities.

2. Where AI may be used

AI may assist with resume parsing, skill and title normalization, semantic search, job and candidate recommendations, profile completeness, duplicate or inconsistency detection, interview transcription and summarization, assessment support, market-range insights, workflow prioritization, fraud detection, and generation of explanations or communications.

TruTalent will identify AI-assisted features in the interface, notice, documentation, or customer materials where the use could reasonably affect a person or materially influence a hiring workflow.

3. What AI outputs are — and are not

Outputs are probabilistic decision-support signals. They may be wrong, incomplete, biased, stale, or based on insufficient context. A match percentage, trust indicator, salary range, verification signal, interview summary, or recommendation is not a fact, guarantee, certification, legal conclusion, or final hiring decision.

4. Human oversight

Customers must ensure meaningful human review before rejecting, excluding, ranking down, changing compensation, or otherwise materially affecting a candidate based on an AI output. Reviewers must be able to understand the principal factors, consider additional evidence, override the output, document material decisions, and handle candidate challenges.

5. Candidate notice and contestability

Where AI materially supports an employment workflow, candidates should receive clear notice of the feature, purpose, data categories, output type, role of human review, retention, and available alternatives where required. Candidates may report inaccuracies, provide context, request correction, and seek human review through privacy@trutalent.ai.

6. Restricted and prohibited AI uses

TruTalent and customers must not use the Service to:

• make fully automated final hiring or firing decisions without lawful safeguards;
• infer race, caste, religion, political opinion, sexual orientation, disability, health, trade-union membership, or other protected traits from unrelated data;
• perform emotion recognition in the workplace or recruitment where prohibited or scientifically unsupported;
• rank people based on social behavior unrelated to job requirements;
• use facial, voice, or biometric analysis without a separate lawful, necessary, validated, and disclosed workflow;
• intentionally proxy for protected characteristics;
• manipulate or deceive candidates; or
• claim that an output is unbiased, objective, or guaranteed.

7. Data and model governance

TruTalent maintains documented feature purposes, data provenance, quality controls, evaluation datasets, model and prompt versions, performance thresholds, error analysis, bias and disparate-impact testing, security review, change approval, monitoring, and rollback criteria proportionate to risk.

Customer Content from paid enterprise workspaces will not be used to train generally available foundation models unless the customer expressly opts in through a written and revocable arrangement. Aggregated or de-identified telemetry may be used to improve reliability and safety where permitted.

8. Transparency information

For material AI features, TruTalent should maintain a model or system card describing: intended use; prohibited use; input and output categories; principal factors; known limitations; evaluation approach; human-oversight requirement; data retention; applicable model providers; and change history. Trade secrets and security-sensitive details may be withheld while still providing meaningful information.

9. Fairness and accessibility

TruTalent tests for unjustified disparities across relevant groups where lawful data is available, evaluates job-relatedness, supports reasonable accommodation and accessible alternatives, and investigates reported harms. Customers are responsible for validating their criteria, job descriptions, and local legal obligations.

10. External model providers

TruTalent may use third-party AI providers under contractual controls covering permitted processing, security, retention, confidentiality, subprocessors, and training restrictions. Provider names and material purposes should be included in the subprocessor register or customer documentation.

11. Output disclosure language

For material external use, use:

“This information was generated or assisted by TruTalent AI. It is decision-support information, may contain errors, and requires human review. It must not be treated as the sole basis for an employment decision.”

12. Reporting and review

Report inaccurate, discriminatory, unsafe, or misleading outputs to privacy@trutalent.ai. TruTalent may pause a feature, correct data, investigate logs, notify affected customers, retrain or replace components, and document remediation. This policy is reviewed at least annually and before material high-risk AI changes.