NORBERT BIEDRZYCKI

Who Owns AI Risk? The Missing Link Between Technology and Corporate Governance

Who Owns AI Risk? The Missing Link Between Technology and Corporate Governance

A year ago Zillow shut down Zillow Offers, its business of buying homes directly using algorithmic price estimates. The models had not anticipated how quickly the housing market would change, the company found itself holding thousands of homes worth less than it had paid, and it announced it would cut around a quarter of its workforce. An algorithmic misjudgment became a strategic failure, a large write down and a story about governance.

This year, regulators in several European countries fined Clearview AI for scraping billions of facial images without consent. In August the US National Institute of Standards and Technology published the second draft of its AI Risk Management Framework, with a final version expected early next year.

Different events, one common question. When AI goes wrong, who in the organization owns the risk?

The current answer is nobody in particular

In many companies AI risk sits in the technology function by default. Data scientists build models, IT deploys them, and governance, where it exists, is a technical review.

But look at what AI risk actually consists of. Discrimination is a legal and reputational risk. Privacy violations are a regulatory risk. Model errors can be financial risks, as Zillow shows. Security vulnerabilities are operational risks. Loss of customer trust is a commercial risk.

None of these are primarily technical. All of them land on the board’s agenda eventually, usually after the fact.

What NIST gets right

The NIST framework is voluntary and deliberately general. Its value lies in its structure. It organizes AI risk management into four functions: govern, map, measure and manage. And it places governance first, as a cross cutting function that defines culture, accountability and roles across the whole organization.

That is the key insight. AI risk management is not a model validation step. It is an organizational system that needs owners in legal, compliance, security, data, business operations and the executive team.

A practical model

In my experience with regulated organizations, a workable structure has three lines, much like financial risk.

Business owners are accountable for the AI systems in their processes, including outcomes, not just performance.

A central function, often combining risk, legal and data expertise, sets standards, maintains an inventory of AI systems and reviews the high risk ones.

Internal audit and the board provide independent oversight and ask whether the system works.

The specific design matters less than the principle. Every material AI system should have a named business owner, and AI risk should appear in the enterprise risk register next to credit, market, operational and cyber risk.

The board’s question

Boards do not need to understand neural networks. They do need to ask a few simple questions. What AI systems make or shape material decisions in this company? Who owns each of them? How would we know if one of them was failing? Most boards today would not get a clear answer.

Leadership question: Should AI risk be managed by the CIO or CTO, or should it become a permanent responsibility of the board and executive committee?

Read next

  1. AI Transparency: Can We Know Whether We Are Interacting With a Human or a Machine?

    Ethical AI series, written in 2026

  2. Human in the Loop: Does AI Need Human Supervision to Remain Responsible?

    Ethical AI series, written in 2026

  3. AI Ethics Washing: Are Corporate Principles Enough?

    Ethical AI series, written in 2026