Defensible AI: Operational Principles for Governing Intelligent Systems

Free download. Pay what you want, including nothing.

Most AI systems work. Far fewer can prove what they did when a regulator, a buyer, or a court asks.

Defensible AI: Operational Principles for Governing Intelligent Systems is an operating doctrine for the people who answer for AI and the people who build it. It rests on one claim: an AI system can work perfectly and still be difficult to defend. Closing that gap is something you build before scrutiny arrives, never something you reconstruct after.

Who it's for

• Founders and small teams: the Startup Prompt finds your gaps in an afternoon.
• Counsel and compliance leads: what regulators actually ask for, with a crosswalk to NIST AI RMF 1.0 and ISO/IEC 42001.
• CTOs and ML leads: the decision record your systems must produce on demand.
• Boards and executives: Ten Board Questions and a maturity score to track over time.

What's inside

• Five principles for building AI systems: Define, Design, Build, Control, Validate.
• Five layers for running them, and the seams where failures actually happen.
• The Governed Decision Record: the evidence you must produce on demand for a single consequential decision.
• A 200-point Maturity Index, a 90-Day Sprint, and an operational checklist.
• Appendix A on the 2026 regulatory picture: the EU AI Act, California, Colorado, Texas, and where enforcement is heading.
• Case studies including Air Canada, Knight Capital, Samsung, CrowdStrike, FTX and Coinbase, and Mobley v. Workday.

Four companion prompts you can run today

Appendix I turns any capable AI model into a working partner: a Startup Prompt for small teams, a Governance Planning Prompt, an Assessment Prompt that scores your system against the Maturity Index, and a Remediation Prompt that closes the gaps the others find. Attach the PDF and your own plans, and you will have a list of your own gaps before you finish Chapter 1.

Two files, ready to download: a 250-page PDF for reading and for use with the companion prompts, and an EPUB for phones and e-readers.

Written by Bernhard D. Kirchner, founder of Syn Data Labs, which helps organizations in regulated markets build AI systems that withstand scrutiny. A former U.S. Army officer and Foreign Area Officer, he holds a JD from Fordham Law and degrees from Columbia and Georgetown.

Pay what you want, including nothing. One hundred percent of anything paid for this digital edition, after platform and payment processing fees, is donated monthly to organizations helping veterans escape homelessness.

General information only; not legal or compliance advice.