Why We Wrote a Book About AI Danger (and What It Taught Us About Safe AI)

Before founding Smith & Gray, I wrote a book called "DAiNGER, DAiNGER!" — a practical guide to AI risks and opportunities. It was not an academic exercise. It was born from years of seeing AI deployed without rigour.

I had designed programmes for federal government agencies, where the stakes are highest. When you are building AI for government, safety is not optional — it is the foundation. Every decision is documented, every risk is assessed, every system has fail-safes.

But in the private sector, I saw a different approach. Companies deploying AI because it was trendy, not because it was safe. AI agents making decisions without human oversight. No audit trails. No rollback capability. No safety review.

The book was my attempt to bridge that gap — to bring the rigour of government programmes to businesses of all sizes. Not to scare people away from AI, but to show them how to use it safely.

The key lesson from writing the book: safe AI is not about saying no to AI. It is about deploying it with the right controls — human-in-the-loop, audit trails, confidence thresholds, rollback capability, and Principal review.

That is what Smith & Gray does. Every agent we deploy has these controls built in. Every deployment is reviewed by the Principal — me. Not because I do not trust my team, but because safety review at the top level is non-negotiable.

The businesses that will win with AI are not the ones that deploy it fastest. They are the ones that deploy it safest. That is what "Safe AI, Done Right" means to us.