The book
Explains the auditability gap
Why AI decisions in regulated settings so often cannot be reconstructed, reviewed, or defended - and what defensibility actually requires.
The book
The Auditability Gap
A book about the hidden risk in AI systems that cannot be reconstructed, reviewed, or defended after the fact.
IEG
By Ian Hafkenschiel, US-trained lawyer, senior software engineer, founder of LexLatam, and creator of Audit Grill-Me.
The Auditability Gap: Non-Reproducible Intelligence in Regulated Systems153 pages - paperback & KindlePublished March 2026
The risk is not only what the AI said. It is whether you can prove how it got there.
Regulated organizations are adopting AI faster than their evidence systems are evolving. Policies, principles, and governance committees are not enough if the organization cannot reconstruct the actual chain of inputs, prompts, sources, model behavior, human review, and retained evidence behind an AI-assisted decision.
What the book covers
Five arguments, built for the people who have to answer for the system.
01
Why non-reproducible AI creates governance and audit exposure
02
Why traditional software audit assumptions break down with generative systems
03
Why model risk is only part of the problem
04
How evidence, provenance, logging, source grounding, and human review shape defensibility
05
Why auditability must be designed into the system, not added afterward
How this connects to Audit Grill-Me
The book makes the argument. The sprint puts it to the test.
The sprint
Tests whether your workflow has it
A fixed-scope evidence readiness assessment that takes one real AI workflow and checks whether its decisions can be reconstructed on demand.
Audit Grill-Me turns the book's thesis into a practical stress test. The book explains the auditability gap. The sprint tests whether one of your workflows has it.
Next step
Read the thesis. Then test your workflow.
The book is not legal advice. The Audit Grill-Me Sprint is an engineering and auditability assessment, not a legal opinion.