# AI Governance Policy — How Talent Pronto Governs Anna > How Talent Pronto governs its AI interviewer: intended use, human oversight, fairness safeguards, accommodations, data privacy, change management, and explainability. Source: https://www.talentpronto.ai/ai-governance-policy --- AI Governance Policy # Anna asks the questions. We answer for how she scores. This policy sets out how Talent Pronto designs, deploys, and oversees artificial intelligence in its hiring platform — including Anna, our AI interviewer, and the structured screening, rubric-based scoring, and candidate communications she handles. It applies to all Talent Pronto personnel and to the production systems that deliver these capabilities. **Owner** Chief Technology Officer **Review cycle** Annual **Formalized under** our SOC 2 readiness program Contents [1Accountability](#accountability) [2Intended use and human oversight](#intended-use) [3Fairness and bias safeguards](#fairness) [4Accessibility and accommodations](#accessibility) [5Data privacy and security](#privacy) [6Change management](#change-management) [7Transparency and explainability](#transparency) [8Regulatory alignment and review](#regulatory) 01 ## Accountability Our Chief Technology Officer is the accountable owner for AI governance, ethics, compliance, and model risk. The owner approves material changes to models and evaluation logic, oversees validation reviews, and conducts the annual review of this policy. Concerns about AI behavior or outcomes — raised by customers, candidates, or personnel through any support channel — are escalated directly to the owner. 02 ## Intended use and human oversight Anna conducts structured first-round screening interviews and scores candidate responses against a job-specific rubric defined with the customer at job setup. Each candidate is evaluated independently against that rubric; candidates are not ranked or compared against one another. **Anna does not make hiring decisions.** Scores, rationale, and transcripts are inputs to a human process, and customers retain full decision authority. Optional automation, such as knockout rules for stated minimum requirements, is configured by the customer and can be fully disabled so that every candidate receives human review before any adverse decision. Prohibited uses Anna is not used to infer or evaluate protected characteristics, to make final employment decisions, to monitor candidates outside the interview itself, or for any purpose other than candidate screening. 03 ## Fairness and bias safeguards - We do not collect demographic or protected-class information in the interview or scoring flow, and no such information is used in scoring. - Scoring instructions are restricted to the job-related criteria in the rubric and explicitly exclude protected characteristics. - Every candidate for a given job receives the same structured interview and is scored against the same rubric under the same evaluation logic. - Scoring behavior is reviewed internally with each material release and at least semiannually, as described in our summary of Anna’s scoring safeguards, validation, and limitations. - We support customer-initiated or independent third-party validation where a customer’s policies or jurisdiction require it. 04 ## Accessibility and accommodations Candidates who request an accommodation or an alternative to the AI interview are routed to the customer’s standard manual process. Candidates are never automatically rejected or withdrawn for non-completion where an accommodation applies, and customers can bypass Anna screening for any individual candidate. We continue to invest in accessibility, consistent with the Americans with Disabilities Act and comparable requirements. 05 ## Data privacy and security Each customer’s data is kept logically separate. **Customer and candidate data are not used to train or fine-tune AI models.** Data is retained and deleted in accordance with customer agreements. Our security program, developed under our SOC 2 readiness effort, includes access controls, secured ATS integrations, a documented incident response plan, and periodic security testing. 06 ## Change management Material changes to models, prompts, or evaluation logic are validated before release and versioned. Customers receive advance notice of material changes that could affect scoring outputs. Routine product updates are communicated through release notes and customer success contacts. 07 ## Transparency and explainability Candidates are informed that they are interacting with an AI interviewer. Every result includes per-question scores and written rationale tied to the rubric, along with the interview transcript, so reviewers can see exactly how a score was reached. 08 ## Regulatory alignment and review We monitor developments in employment and AI law relevant to our customers — including EEOC guidance, state laws governing AI in hiring, and local bias-audit requirements — and we update our product and this policy as requirements evolve. This policy is reviewed at least annually by the owner, and material revisions are communicated to customers. For how these commitments show up in the product, see our [security, compliance, and fair hiring overview](https://www.talentpronto.ai/security-compliance-fair-hiring-page). ## Reviewing us for your vendor assessment? Our compliance team can walk your legal, security, or HR stakeholders through this policy and provide the supporting documentation behind it. [Email compliance@talentpronto.ai](mailto:compliance@talentpronto.ai) [See our compliance overview](https://www.talentpronto.ai/security-compliance-fair-hiring-page)