Controls & Safeguards

Values are necessary. But values don’t scale by themselves.

When technology accelerates, incentives distort, and people are under pressure, the difference between “what we believe” and “what we actually do” becomes the risk.

This page is where we make ethics operational:
controls, safeguards, release gates, and verification that hold up when it’s hardest to hold them.

What we mean by “controls”

A control is a constraint on behavior. It changes what happens when:

timelines tighten
reputational pressure rises
authority dynamics distort decisions
ambiguity is exploited
dependency starts to form

A control can be:

preventive (stop harm before it starts)
detective (surface drift quickly)
corrective (reduce blast radius, restore safety)
directive (gates, approvals, required steps)
recovery (rollback, escalation, kill switch)
governance (ownership, accountability, auditability)

Start here

1) The Controls Library (5–15 minute safeguards)

Deployable, portable controls you can run immediately with your team, and leave behind a trace.

Go to the Controls Library:
https://cwmetz.com/controls-library/

2) Two foundational explainers (PDF)

Manifestos Without Controls Are Just Poetry
https://cwmetz.com/wp-content/uploads/2026/01/manifestos-without-control-are-just-poetry.pdf
The Three Manifestos (Orienting / Internal / Public) — mapped to controls + release gates
https://cwmetz.com/wp-content/uploads/2026/01/the-three-manifestos.pdf

How to use this section

Pick one control that matches a real decision you’re making this week.
Run it once while the decision is live (not in theory).
Save a trace (even one sentence).
Repeat under pressure until it becomes habit.

Want the “enforced” version?

These controls are designed to stand alone.

SpiralWatch™ implements the same logic as a fail-closed assurance layer for human-facing AI: scenario-driven expectations, stop conditions, and audit-ready evidence packs.

A Think Tank for Human Flourishing