Cinderpoint · Responsible-AI Readiness

SAFEMACHINE

A free, offline responsible-AI readiness tool, and the body of work behind it. SAFEMACHINE turns structured questions into a defensible GO, REVIEW, or NO-GO with the reasoning attached. This page is the home of the app, the OUTPACED series, the frameworks it runs, the thirteen failure mechanisms it tests for, the papers, and the worksheets.

Download the app The 13 mechanisms Worksheets
The App
Free · Offline · Desktop
SAFEMACHINE
A deterministic, rule-based readiness check that runs entirely on your device. No accounts, no telemetry, no cloud. Three ways in: a 10-minute individual self-check, a full organization review, or a runtime-AI review. Every decision carries a versioned paper trail you can export and reproduce.
Why it stays free
Ethical Monetization
The app is free. The commitment behind it, and how to verify every word of it, is published in full.
The Frameworks

Five frameworks make up the SAFEMACHINE method. BASE establishes readiness, SAFEARC produces the decision, STABLE handles incidents, CARG governs adaptive runtime AI, and OUTPACED diagnoses the failure mechanisms.

Readiness
BASE
Eight readiness questions establish whether an organization is ready to proceed before the governance review begins.
Decision engine
SAFEARC
A seven-pillar architecture (Scan, Align, Filter, Evaluate, Assign, Renew, Contain) that produces the GO / REVIEW / NO-GO and the hard stops.
Incident response
STABLE
A seven-step containment workflow that opens when an AI system causes harm. Tracks severity, response, and closure.
Runtime governance
CARG
Six-component runtime governance for self-modifying AI, including the SMC-4 x HRP-4 restricted-deployment hard stop.
Failure diagnostic
OUTPACED
The thirteen mechanisms by which oversight falls behind a deployed AI system. The diagnostic at the heart of the book.
OUTPACED · The 13 Failure Mechanisms

The thirteen mechanisms by which an organization's oversight falls behind the AI it has deployed. Each is a canonical concept page drawn from Outpaced by Waydell D. Carvalho.

Mechanism 1
Illusion of Validity
Stable tracked metrics get read as proof the system is safe.
Mechanism 2
Illusion of Control
Owning a system is mistaken for being able to steer its decisions.
Mechanism 3
Ceremonial Compliance
Governance that runs on schedule and produces records but changes nothing.
Mechanism 4
Pacing Problem
Oversight can only constrain the system long after it has moved on.
Mechanism 5
Alert Fatigue
The signal that should interrupt a failure is lost in the noise of alerts.
Mechanism 6
Normalization of Deviance
Operating outside the original specification quietly becomes the new standard.
Mechanism 7
Post-hoc Rationalization
The explanation is built after the decision, not from what produced it.
Mechanism 8
Surrogation
The proxy metric quietly replaces the outcome it was meant to track.
Mechanism 9
Problem of Many Hands
Everyone owns a piece of the system; no one owns the whole.
Mechanism 10
Coordination Neglect
Many overseers act separately; no one combines them into one constraint.
Mechanism 11
Threat Rigidity
Evidence the premise is failing triggers doubling down, not revision.
Mechanism 12
Moral Licensing
A stated ethical identity stands in for scrutiny of specific decisions.
Mechanism 13
Cascading Failures
Accumulated hidden state turns small faults into fast-moving chains.
The Series
Book
Outpaced
The thirteen ways organizations fall behind the AI they deploy, and the diagnostic that names them. The foundation of the SAFEMACHINE method.
Papers
Working paper · Zenodo
SAFEARC
A seven-pillar operational architecture for socio-technical AI governance. The methodology behind the decision engine.
Preprint · Zenodo
Runtime Governance (CARG)
A framework for continuous oversight of self-modifying and continuously-adapting AI systems.
Index
All Papers
Every working paper and preprint behind the Cinderpoint frameworks.
Worksheets

The same worksheets the app runs, free to download and use on paper. Each carries the Cinderpoint brand, the author, and the disclaimer. Word and PDF are the same content; the version is printed inside each file.

These worksheets support governance review. They are not legal, compliance, or professional advice. The links below always point to the current version.

BASE
Organizational readiness baseline. Complete before the SAFEARC review.
WordPDF
SAFEARC
The seven-pillar governance pipeline. Produces the GO / REVIEW / NO-GO.
WordPDF
STABLE
The seven-step incident-response and containment workflow.
WordPDF
OUTPACED
The thirteen failure mechanisms. Individual self-assessment and the deeper organizational diagnostic.
WordPDF
CARG
Six-component runtime governance for self-modifying AI, with the restricted-deployment hard stop.
WordPDF