The CISO role is expanding. What began as responsibility for information security has progressively extended to data privacy, supply chain risk, and now — with increasing urgency — AI governance. In regulated industries, CISOs are frequently the executive most operationally equipped to own the AI governance function, and increasingly they are being asked to do so.
Why AI Governance Is Landing on the CISO's Desk
The answer is structural. AI governance requires the same cross-functional authority, risk management discipline, audit infrastructure expertise, and regulatory literacy that effective CISOs already exercise. The CISO already owns the frameworks (ISO 27001, NIST CSF) that ISO 42001 and NIST AI RMF are modelled on. The CISO already runs the audit and evidence function that AI compliance requires. And the CISO already has the board relationships and incident response infrastructure that AI risk management demands.
This does not mean AI governance is purely a security function. It has dimensions — ethics, employment law, consumer protection, sector regulation — that require input from legal, compliance, HR, and business leadership. But the CISO is well-positioned to be the integrating function: the owner who pulls cross-functional input together into a coherent AI governance programme.
The AI Regulatory Landscape for Regulated Industries
CISOs in regulated industries face a layered AI regulatory landscape where general AI legislation sits on top of sector-specific obligations. Understanding this layering is essential for designing a governance programme that is efficient and comprehensive.
Financial services: The EU AI Act's high-risk AI provisions directly cover credit scoring, insurance risk assessment, and recruitment — activities core to most financial services firms. Layered on top: EBA guidelines on AI in financial services, the FSB's framework on AI in financial stability, and national prudential regulator expectations (FCA in the UK, BaFin in Germany). SR 11-7 model risk management guidance from the Federal Reserve applies to US bank holding companies and has been interpreted by most major banks as covering AI models.
Healthcare: AI in medical devices is classified as high-risk under the EU AI Act and subject to MDR 2017/745 conformity requirements. HIPAA imposes obligations on AI systems that process protected health information. FDA guidance on AI-based software as a medical device (SaMD) applies to US-market deployments.
Insurance: State insurance commissioners in the US have issued AI governance guidance emphasising algorithmic accountability, with NAIC model bulletin providing a reference framework. In the EU, EIOPA has published guidelines on AI governance in insurance that complement the AI Act.
Building the AI Governance Programme
The CISO's AI governance programme should be built on four pillars: Inventory and classification — every AI system in production catalogued, classified by risk tier, and mapped to applicable regulatory requirements. This is the AI asset register, the AI equivalent of the information asset register the CISO already maintains. Risk management — AI-specific risk assessments integrated into the existing risk management framework, with AI risks reportable alongside other enterprise risks. Audit and evidence — the audit trail infrastructure that produces tamper-evident records of AI system operation, enabling post-hoc verification and regulatory response. Incident response — AI-specific scenarios integrated into the cyber incident response plan, with defined escalation paths for AI performance failures, bias incidents, and adversarial attacks.
The Audit Trail as Security Infrastructure
CISOs understand audit logs. The AI audit trail is a direct extension of the security logging infrastructure the CISO already operates — but with specific requirements that differ from traditional security logging. AI decision logs must be tamper-evident, not just tamper-resistant. They must capture decision-level semantics (what did the model decide, on what inputs, with what confidence) rather than just event-level metadata. And they must be retained for periods — 7–10 years — that far exceed standard security log retention cycles. The good news: the cryptographic integrity techniques that underpin security log integrity (hash chaining, WORM storage, third-party attestation) transfer directly to AI audit trail requirements.
Reporting AI Risk to the Board
CISOs in regulated industries increasingly need to report AI risk to audit committees and boards. The most effective board AI risk reporting frames AI risk in business terms: regulatory exposure (which regulatory requirements apply, what the penalty exposure is, where the gaps are), reputational risk (what AI decisions could attract public or regulatory scrutiny), and operational risk (what AI system failures could disrupt core business processes). Pair the risk register with evidence of controls — showing not just that risks have been identified, but that they are being actively managed and that the organisation has the evidence to demonstrate this to regulators.