AI Governance for Small Businesses: Where to Start
October 2026 · 4 min read
Most AI governance guidance is written for enterprises with compliance teams. If you're a small business using AI tools — chatbots, copilots, automated decision-making — you need something simpler. Start with three things:
- Inventory. Write down every AI tool and model your business uses, including the ones employees adopted on their own. You can't govern what you can't see.
- Risk tiering. Sort each use case into low, medium, or high risk. A marketing copy generator is not the same as a tool that screens job applicants.
- One policy. A single page: who approves new AI tools, what data they can touch, and who is accountable. That's your foundation.
Where small businesses usually stand
Takeaway: you don't need a 200-page framework on day one. Inventory, risk tiers, one policy — then run a gap scan to see what's next.
ISO 42001 vs NIST AI RMF: What's the Difference?
October 2026 · 5 min read
Two names come up in every AI governance conversation: ISO/IEC 42001 and the NIST AI Risk Management Framework. They solve different problems.
- ISO 42001 is a management system standard — like ISO 27001, but for AI. It tells you how to build an AI management system: define scope, assign roles, run the lifecycle from design to decommissioning. It's auditable and certifiable.
- NIST AI RMF is a risk framework. It gives you four functions — Govern, Map, Measure, Manage — and asks how well you perform each. It's voluntary, flexible, and excellent for prioritization.
In practice they complement each other: use NIST AI RMF to understand your risks, use ISO 42001 to run the system that controls them. Krypton GRC maps controls to both, so one scan covers you either way.
Takeaway: RMF for thinking, 42001 for operating. Start with whichever matches your next conversation — a regulator or an enterprise customer.
The 5 AI Governance Gaps We See Most Often
October 2026 · 3 min read
Across early scans, the same five gaps show up again and again:
- 1. No AI inventory. Teams can't list what models and tools are in use.
- 2. Untested models. No evaluation before deployment, no red-teaming.
- 3. No risk assessments. High-impact use cases running without documented review.
- 4. Shadow AI. Employees pasting company data into public AI tools.
- 5. Missing policies. No written rules on acceptable AI use.
How often each gap appears in initial scans
Takeaway: if any of these sound familiar, that's your starting list. A Krypton readiness scan turns them into a prioritized plan.