Glossary
What is AI governance?
In short
AI governance is the system of rules, roles and checks by which a company decides which AI systems to launch, how to watch their risks and who is responsible for the result.
Business and adoption · AI governance
MMOOZZGG.com editorsChecked: 1 min read
For a child
Imagine a summer camp gets pedal boats. The counselors decide who can ride them, from what age, in which life jackets and who watches from the shore. Every morning they check that all the boats are in good shape. Without such rules, the fun would end quickly. Companies agree on rules for their smart programs in the same way.
In plain words
When a company has dozens of AI services, it needs order: who approves a launch, by what rules, how to find out that a system makes mistakes. The company keeps a list of all such systems, assigns an owner to each, assesses the risk before launch and checks the work afterward. There is a clear procedure for when a model gives a harmful or wrong answer. It is like traffic rules: the cars differ and the drivers differ, but the signs, licenses and inspections are the same for everyone.
For business
Without this system, a company learns about a problem last: from a customer complaint, a regulator’s check or a leak. The rules protect revenue and reputation and even speed up launches, because every project has a clear approval path. Costs depend on scale: a small business needs a usage policy, a register of services and a person in charge, while a large company needs a committee, vendor assessment and regular audits. The risks this work covers: personal data handled outside the rules of 152-FZ, the Russian personal data law, decisions based on unchecked answers, dependence on one vendor and breaches of others’ rights to content. Before you start, check whether you have a full list of the AI systems your staff already use. Such an inventory is also where a review of AI risks in a company usually begins.
For a specialist
AI governance is the set of policies, processes, roles and technical controls that keep AI systems within the organization’s goals, the law and acceptable risk across the whole life cycle. The core document, the NIST AI Risk Management Framework 1.0, was published on 26 January 2023 for voluntary use and splits the work into four functions: Govern, Map, Measure and Manage. Govern covers documenting legal requirements, mechanisms for an inventory of AI systems, assigning roles and responsibility, staff training and policies on third-party risk, including vendors of models and data. In July 2024, NIST released a profile for generative AI (NIST AI 600–1). Practical elements include a register of systems and use cases, risk classification, an impact assessment before launch, vendor checks, monitoring of quality and drift, an incident log and a human in the loop for decisions where an error is costly. The foundation is data governance.
Example
An insurance company found that its departments used nine different AI services. It set up a register, assigned an owner to each service and made a rule: a new tool is launched only after a risk assessment. Two services that stored queries abroad were switched off.
Related terms
Questions and answers
Are AI governance and managing a business with AI the same thing?
Sources
Not sure whether your company needs this?
AI solutions map