> ## Documentation Index
> Fetch the complete documentation index at: https://www.docs.forgeshipyard.com/llms.txt
> Use this file to discover all available pages before exploring further.

# What Forge Will Not Do

> The limits we deliberately build in, so you can trust the system.

export const VerificationLadder = () => <Tooltip tip="The ordered sequence of checks that runs on every AI output before it can become an authoritative decision, ending in human sign-off when required.">verification ladder</Tooltip>;

<Note>
  **In plain terms:** Forge can retrieve, compare, draft, flag, and propose. It
  **cannot** be the design authority. Any decision that touches class rules, Coast
  Guard rules, or safety always passes through a human. This is built into the
  structure, not a setting that can be switched off.
</Note>

## The one rule that governs everything

> AI proposes; humans decide; the gate is structural.

The AI is allowed to do the legwork: pull up similar prior designs, compare
options, draft an assessment, flag a conflict, suggest a path. It is **not**
allowed to be the final word on anything that matters. Decisions touching
classification society rules, USCG compliance, or safety always require a human
to approve them before they become authoritative.

This is not a configuration setting a yard could accidentally disable. It is baked
into the <VerificationLadder /> every output has to pass through.

## Where humans stay in charge

The system is intentionally designed so that humans remain the authority on:

<CardGroup cols={2}>
  <Card title="Class and USCG decisions" icon="scale-balanced">
    Every decision touching classification or Coast Guard rules requires human
    approval, by structure.
  </Card>

  <Card title="What the system knows about" icon="sitemap">
    Changes to the underlying model of what kinds of things exist are governed by
    humans, never by the AI.
  </Card>

  <Card title="What the AI is allowed to do" icon="list-check">
    The catalog of permitted AI moves goes through a review checklist before any
    change ships.
  </Card>

  <Card title="Production sign-off" icon="circle-check">
    Every sign-off that a build is ready for production is a human decision.
  </Card>
</CardGroup>

## When the AI is unsure, it stops

Forge is designed to **abstain rather than guess**. If the AI is not confident, or
it cannot find the evidence it needs, it does not make something up. It asks for
more information or routes the question to a person. Refusing to answer when the
evidence is insufficient is treated as a quality behavior, and the system is held
to a high bar for doing it (a target of better than 98%).

## Why build it this way

The point is to make the AI's behavior **boxable, auditable, and reversible**. If a
behavior turns out to be wrong, it gets rolled back like any other versioned
software change, not patched on the fly with a different prompt. That is what makes
a Forge decision defensible in front of a regulator, a customer, or a court.

See [Where Humans Stay in Control](/concepts/human-control) for the full detail,
and [Keeping Your Data Yours](/concepts/keeping-data-yours) for how sensitive
designs are protected.