← back to the archiveCover illustration for “AI governance: who decides how AI is used, and how”
TERM7mo ago·Published · Updated ·by Andy Padia

AI governance: who decides how AI is used, and how

In short: AI governance is the set of policies, roles, controls and evidence that decides how an organisation builds, buys and uses AI and who answers for it.

AI governance is the system of policies, roles, processes, technical controls and evidence that decides how AI is used in an organisation: which AI to build, buy and deploy, how to manage its risks, and who is accountable when it fails. Frameworks give it structure, such as the voluntary NIST AI Risk Management Framework, released on 26 January 2023.

AI governance in plain words

These are the house rules for AI: who can use what, for which jobs, with what checks, and who answers for it when something goes wrong.

How it works

Governance spans ethics and law, but day to day it is mostly operational. A working programme, in my view, has six parts:

  1. An inventory of AI systems and their owners.
  2. Risk classification and an approval path proportionate to risk.
  3. Technical controls: access, logging, gateways, guardrails.
  4. Evidence: evals, red-team results, model cards and AI-BOMs.
  5. Monitoring and incident response.
  6. Training and acceptable-use policies for staff.

Frameworks supply the structure. NIST's framework is organised around four functions, Govern, Map, Measure and Manage, and gained a Generative AI Profile on 26 July 2024; its page, read in November 2026, says version 1.0 is being revised. The OECD AI Principles, adopted in 2019 and updated in May 2024, set values such as transparency, safety and accountability. Standards such as ISO/IEC 42001 and laws such as the EU AI Act sit alongside them.

Example

As an illustration, a company lets any team experiment with approved models through a gateway, but requires a lightweight review before customer-facing launch, and a full assessment with human oversight for uses touching hiring, credit or health.

Where you'll meet it

  • AI councils and responsible-AI offices
  • Procurement and vendor-risk processes
  • Board and audit committee agendas
  • Platform engineering, where governance becomes enforceable controls

What it changes for you

For leaders, my view is that governance decides speed: clear, proportionate rules let teams move faster than vague ones. For engineers, it shows up as gateways, logging requirements and release gates. For employees, it defines which tools they may use with which data.

AI governance vs AI ethics

AI ethics asks what is right. AI governance turns those answers, plus law and business risk, into processes and controls that actually run.

Common mistake

My view: governance that only says no. When approved paths are slow or missing, people use AI anyway on personal accounts. Good governance provides a fast, safe default path.

Shadow AI · NIST AI RMF · ISO/IEC 42001 · EU AI Act · AI gateway

From the archive: Voluntary AI frameworks are regulatory signals

Good AI governance gives teams a fast, safe default path, because rules that only say no get bypassed.

#field-guide#governance#enterprise-ai#risk#leadership

Sources

  1. nist.gov/itl/ai-risk-management-framework
  2. oecd.ai/en/ai-principles
← older drop
LoRA: fine-tuning a big model by training a small add-on
newer drop →
Prompt engineering: writing instructions models follow well

related drops

explore all 557 drops →
← back to the archiveday 175