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

# Posture

> See what an AI system's risk rating asks of it, which requirements it meets, and close each gap from its Overview.

Posture answers one question an auditor will ask: does each AI system have what its risk calls for? The higher the [rating](/ai-systems/risk-rating.md), the more it needs. Altrum checks every requirement for you, shows the gaps on the AI system's **Overview**, and offers the action that closes each one.

## Requirements per level

| Requirement | Low | Medium | High | Very high |
| - | - | - | - | - |
| Owner is an active member | Yes | Yes | Yes | Yes |
| Risk profile answered | Yes | Yes | Yes | Yes |
| A submitted assessment within | | 12 months | 6 months | 3 months |
| Guardrails live | | Yes | Yes | Yes |
| Guardrails check replies | | | Yes | Yes |
| Every identified risk covered | | | | Yes |
| Budget set on the AI system | | Yes | Yes | Yes |
| Content logging on | | | | Yes |

Guardrails check replies when at least one live guardrail checks what the model answers. A risk is covered as on the **Risk** tab: see [Residual rating](/ai-systems/risk-rating.md#residual-rating).

The budget and content logging rows apply only when your organization has the [AI Gateway](/ai-gateway/overview.md). Content logging counts as on when the AI system's own switch is **On**, or **Follow the organization** with the organization's switched on.

## The score

The score is the requirements met divided by the requirements that apply, as a whole percentage. An AI system **meets its requirements** at 100%. The meter reads **Good** from 80%, **Needs work** from 50% and **At risk** below.

When the AI Gateway does not answer, its rows say **Could not be checked** with **Try again**. They count as not met until they can be checked.

Draft, In review, Approved and Live AI systems have a posture, so you see what going live needs before you ask. A Retired AI system has none.

## Close a gap

**Overview** lists the gaps first, each with the action that closes it. The met requirements are folded under **N met**.

| Gap | Action | Where it goes |
| - | - | - |
| Owner is an active member | **Change owner** | The **Edit** dialog |
| Risk profile answered | **Answer it** | The **Edit risk profile** dialog |
| Assessment | **Start assessment** | A new [assessment](/assessments/run.md) |
| Guardrails live, Guardrails check replies | **Open guardrails** | The **Guardrails** tab |
| Every identified risk covered | **Open Risk tab** | The **Risk** tab |
| Budget set | **Set a budget** | The **AI Gateway** tab, **Limits** |
| Content logging on | **Turn it on** | The **AI Gateway** tab, **Content logging** |

You see an action only when your role can take it.

## Reassessment

At Medium and above, the **Assessment** card on **Overview** says when the next assessment is due: the interval for the level, counted from the last submitted assessment. Once that date passes it says **Overdue**, the requirement becomes a gap, and **Start assessment** is the main button.

The go-live checks ask for a recent assessment at High and above only. See [Lifecycle](/ai-systems/lifecycle.md#the-go-live-checks).

## Find AI systems with gaps

In **AI systems**, the **Posture** column shows each score as a bar, **Meets** at 100% and a dash for a Retired AI system. Filter by **Posture**: **Has gaps** keeps those below 100%, **Meets** those at it.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.