Skip to content

Content Library

Back to Articles

Article June 2, 2026

AI Governance Fails When Risk Arrives Too Late

AI Governance Fails When Risk Arrives Too Late

Artificial intelligence governance must become operational, cross-functional, and embedded from the earliest stages of development if firms are to manage the growing risks posed by AI and generative AI. Speakers argued that model risk professionals are un
  • AI governance must move beyond policies and become embedded in day-to-day operations
  • Generative AI introduces new risks around opacity, accountability, bias, and data quality
  • Regulatory expectations are evolving but remain fragmented across jurisdictions
  • Cross-functional governance structures are essential for effective oversight
  • AI inventories and lifecycle risk assessments help organizations identify and prioritize risks
  • Risk assessments should begin during ideation rather than after deployment
  • Model risk teams possess skills that position them to lead AI governance efforts
  • Fairness, transparency, and accountability must remain central as AI adoption accelerates 

For organizations rushing to deploy artificial intelligence, the greatest risk may not be the technology itself, but waiting too long to govern it.

That was the central argument put forward by a founder of a consultancy specializing in model risk management frameworks, who warned that firms can no longer afford to treat AI governance as a compliance exercise performed after development is complete.

Instead, governance must become operational, embedded, and active throughout the entire AI lifecycle.

Speaking during CeFPro’s Advanced Model Risk conference, the executive argued that traditional model risk management frameworks provide an important foundation, but are no longer sufficient to address the complexity, speed, and opacity associated with modern AI and generative AI applications.

“AI governance, particularly in the insurance industry, must grow up fast,” she said.

The speaker described how organizations increasingly rely on AI systems that influence underwriting decisions, claims management, customer interactions, and internal operations.

While the technologies continue to evolve rapidly, she argued that the core challenge remains unchanged: ensuring that systems are reliable, transparent, explainable, and fair.

“At the heart of this challenge is risk,” she said, warning that traditional concerns around model explainability and unintended bias become amplified when AI systems operate at scale.

The executive highlighted a growing list of risks associated with both traditional AI and generative AI applications, including complexity, lack of transparency, bias, privacy concerns, legal exposure, reputational damage, and potential harm to consumers.

She noted that generative AI compounds these challenges by introducing vast external datasets, decentralized experimentation, and limited visibility into how outputs are produced.

Regulators have begun responding. The speaker pointed to guidance emerging from insurance regulators including the National Association of Insurance Commissioners, Colorado regulators, and the New York Department of Financial Services. Yet she cautioned against viewing regulation as the primary solution.

“These regulatory developments are welcome but uneven,” she explained. “They signal intent, not full alignment.”

Instead, she argued that organizations must proactively establish governance structures capable of adapting as regulatory expectations continue to evolve.

Central to that effort is operationalizing governance rather than simply documenting it.

“The question isn’t just about compliance,” she said. “It’s about creating systems of governance that are functional, effective, and embedded into the DNA of how companies operate.”

The executive advocated for enterprise-wide AI governance frameworks built around transparency, accountability, privacy, fairness, and security. However, she stressed that principles alone are insufficient without supporting structures.

Those structures should include cross-functional governance committees, clearly defined roles and responsibilities, escalation processes, and strong collaboration among risk, technology, legal, compliance, data science, and business teams.

The objective is to eliminate silos and ensure AI risks are viewed through multiple perspectives simultaneously.

A Chief AI and Data Officer at an international payments processing platform, who introduced the session, said one of the most powerful themes was the speaker’s insistence that risk management teams work alongside the business rather than acting as a separate control function.

“What’s required is a cross-functional, proactive effort that starts long before a use case goes live,” he observed.

The executive described this approach as a mandate for closer collaboration between innovation and oversight functions.

“The days of siloed oversight are numbered,” he said.

A major focus of the discussion centered on AI inventories and risk assessments.

Events you may be interested in

View all of our upcoming events

Sign Up to Our Newsletter

Sign up

Connect Plus

Unlock more with Connect Plus membership

This content is available to Connect Plus members.