AI Governance Professional Profiles: Assurance

Assurance is an increasingly important function that foundational model developers and policy experts alike are identifying as a need. These roles are being identified in the broader market as necessary accelerants to increase AI adoption. It is anticipated that this category will likely mature the most in the next version of the skills framework with greater clarity around assessment and audit responsibilities between the second and third lines of defense within organizations, as well as a clearer understanding of requirements for external evaluation and audit functions.  

Additionally, efforts like risk assessment and technical evaluation profiles could be considered assurance functions. We will continue to monitor and analyze how these roles evolve.  

At present, however, it important to make a distinction between evaluators who build and perform technical evaluations and those who expand conventional technology assessments focused on management system reviews.

These assurance functions require different degrees of knowledge. Those working on technical assessments and audits need to understand AI architecture, model training and data processing, whereas management systems assessors and auditors need greater familiarity with internal and external policy requirements and appropriate process and decision-making structures. Note that these assurance roles may either perform the assessments, audits and evaluations or support internal teams who work to prepare for such evaluations. Preparing for these reviews requires the same level of skills and knowledge.

Technical AI assurance leads  

Alternative titles: AI internal auditor, Technical internal auditor (AI systems), AI assurance auditor, Model and AI systems auditor

Summary statement: Provides independent technical assurance over AI systems, controls and governance processes to ensure compliance with internal policies, regulatory expectations and responsible AI requirements.

Key purpose: Delivers objective assurance that AI systems and related controls are designed and operating effectively across the AI life cycle.