Auditing AI Systems in Critical Sectors

AI is moving from simple human-to-system interaction to agentic AI, where machines interact with other machines, take actions, exchange data, and influence business decisions.This creates a major challenge for cybersecurity, GRC, audit, compliance, and assurance professionals:

Priyank Sonihttps://www.linkedin.com/in/priyank-soni-iima/How do you audit an AI system when you cannot fully see the model, logic, architecture, or internal decision-making process?In this podcast, Prabh speaks with Priyank Soni, a cybersecurity, AI/ML, digital trust, governance, and assurance leader, about auditing AI systems in critical sectors.Priyank explains why traditional audit approaches are not enough when AI systems become black boxes, especially in sectors where trust, safety, compliance, and accountability matter.In this episode, we discuss:Why AI audit is becoming important in critical sectorsHow AI is moving toward agentic and machine-to-machine interactionsWhy zero-trust architecture must evolve for AI agentsWhy AI systems require stronger documentation and evidenceHow ISO/IEC 42001 is shaping AI governance and audit expectationsWhy auditors must understand data flow, model behavior, and system boundariesHow to audit black box AI systemsWhy AI audits should start with inventory and classificationWhy data mapping is critical for AI assuranceHow to assess AI vendor and supply chain riskWhy SBOM and ABOM may become important for AI system auditsHow to handle trade secret challenges when vendors do not share model detailsWhy auditor, vendor, and internal team collaboration is requiredWhat performance metrics auditors should reviewHow to test bias, fairness, and reliabilityWhy adversarial input testing mattersWhy human oversight must be validated, not just mentioned in policyWhy AI systems need continuous monitoring after deploymentWhy AI audits may need to happen more frequently than traditional IT auditsPriyank also explains that AI auditing is still in an early maturity phase. Many organizations, vendors, and auditors are learning together. That makes documentation, risk assessment, monitoring, and evidence collection even more important.#AIAudit #AIGovernance #ISO42001 #AISecurity #CyberSecurity #GRC #DigitalTrust #ResponsibleAI #CriticalInfrastructure #VendorRiskManagement #CoffeeWithPrabh