The information security management system supporting the Talent Intelligence Platform is certified against ISO/IEC 27001:2022. The certified scope covers relevant operational, engineering, product, support, legal, finance, human resources, and security functions.

The privacy information management system supporting the Talent Intelligence Platform is certified against ISO/IEC 27701:2019. The scope explicitly includes responsibilities for processing personally identifiable information.

The artificial intelligence management system supporting the Talent Intelligence Platform is certified against ISO/IEC 42001:2023. The scope includes the functions responsible for developing, operating, supporting, and governing AI systems.

Customer system integrations are secured during transmission. This reduces the risk of unauthorized interception while candidate and recruiting information moves between systems.

Customer integration data is secured when stored in databases and files. This helps protect information if unauthorized access to underlying storage is attempted.

Raw resumes used in model training are stripped of direct identifiers and transformed into feature representations without identifiers. The documented analysis reports that nearly all training vectors have multiple close neighbors, reducing the risk of singling out an individual.

Customer data entered or uploaded to the platform is managed by the customer. This supports customer accountability and control over information processed through the service.

Customer data is retained according to contractual commitments and applicable internal retention and deletion policies. This provides a defined basis for managing how long customer information is kept.

Administrators have user-friendly capabilities to locate and take action on candidate data, including deletion of candidate profiles. This supports privacy requests and customer-managed lifecycle processes.

Integrations with customer systems enable customers to manage control over data synchronization. This helps customers govern how information is exchanged with connected systems.

An AI management system is implemented and certified against ISO/IEC 42001:2023. This establishes a formal governance structure for AI-related activities within the documented scope.

Responsible AI practices are supported by cross-functional stakeholders, an AI Ethics Council, and external consultants. These governance participants contribute to the responsible design and deployment of AI systems.

The matching model and AI Interviewer are designed to support human decision-makers rather than make hiring decisions automatically. Recruiters or customer administrators remain responsible for reviewing information and determining next steps.

Users receive understandable explanations of match scores and their relationship to job requirements, including skills, experience, work, and title relevance. This helps customers evaluate AI-supported recommendations and communicate their basis to stakeholders.

The matching model does not use protected characteristics such as gender, age, or ethnicity as model features. This limits the direct use of sensitive attributes in candidate matching.

Candidate masking can remove sensitive information from resumes before recruiter review, including names, gender indicators, race or ethnicity, disability, religion, and contact information. This helps focus evaluation on job-related skills and experience.

The organization measures group-level outcomes using fairness metrics such as impact ratios, selection rates, and statistical performance measures. These analyses support identification of disparate outcomes across gender, race, and intersectional groups.

Perturbation testing compares model scores for original and modified resumes where demographic or other sensitive signals are changed. This provides a documented method for testing whether those changes materially affect matching outcomes.

Annual independent third-party bias audits are conducted for the matching model, with summary information released to customers. This provides external review of fairness and supports customer compliance activities.

Model updates undergo quantitative and qualitative testing, including blind evaluations of model outputs, before release. This helps identify performance or fairness issues before changes reach customers.

Production models are continuously monitored using measures such as prediction behavior, latency, accuracy, and fairness-related metrics. Monitoring can be analyzed across accounts, regions, roles, and other applicable groupings.

Alerts are configured to identify potential model drift or degradation when monitored values cross defined thresholds. This supports timely investigation and response to changes in production behavior.

When critical degradation alerts occur, the engineering organization evaluates the root cause and remediates the issue. Model launch checks can also stop releases when defined gender-difference thresholds are exceeded.

Production models are evaluated regularly against continually growing datasets curated through human-in-the-loop review. This provides an ongoing reference for assessing model behavior.

An independent security firm performed third-party testing of the web application and supporting infrastructure. The engagement evaluated areas including authentication, session tracking, authorization, file handling, and input validation.

An independent security firm performed third-party testing of the external network to identify vulnerabilities that could expose sensitive business data or workflows. The assessment reported no vulnerabilities during its testing period.

The documented testing methodology prioritizes vulnerabilities most likely to be exploited by criminals and combines manual analysis with automated techniques where they improve testing efficiency. The approach is aligned with recognized security and compliance frameworks.

Customer instances are provisioned as private secure areas when the contract begins. This supports separation of customer environments within the platform.

The platform hosting the matching model is managed by the organization. Centralized management supports consistent operation of the platform environment and its associated security processes.

Customer-originated data entered or uploaded by authorized customer users remains managed by the customer. This clarifies customer control and responsibilities for information handled through the platform.

Consent for AI interviews and required video and audio collection is obtained in advance and stored. Consent for recordings can also be collected and made available to customers when required.

Candidates receive advance notification when an AI interview is scheduled, along with interview guidance, frequently asked questions, and an introductory video. These measures help candidates understand the process before participating.

Customers can provide candidates with the option to interview with a human interviewer. This gives organizations a documented alternative to an AI-mediated interview.

Candidates can request accessibility-related accommodations through the documented interview support options. This helps reduce barriers to participation in the hiring process.

The AI Interviewer measures and evaluates whether interview questions avoid requesting sensitive personal information, including health, religion, race, family status, and financial information. This supports privacy-conscious interviewing.

AI-generated interview feedback is evaluated for factual grounding, traceability to transcript evidence, and completeness. These safeguards help customers distinguish supported observations from unsupported conclusions.

Customer administrators can review interview audio, transcripts, summaries, and completed feedback forms before making a decision. Candidates do not automatically advance to later stages based solely on the AI interview.

The AI Interviewer is evaluated for attempts to bypass its guidelines, manipulate its behavior, or induce unauthorized responses. This testing addresses foreseeable misuse and supports safer operation.