> ## Documentation Index > Fetch the complete documentation index at: https://dev-docs.altrum.ai/llms.txt > Use this file to discover all available pages before exploring further. # Lifecycle > Take an AI system from Draft to Live with an approval, pause it, retire it, and see what the go-live checks need at each risk level. Every AI system is in one stage: **Draft**, **In review**, **Approved**, **Live** or **Retired**. Only a **Live** AI system takes traffic: the AI Gateway refuses its keys in every other stage. A new AI system starts in **Draft**. The page header shows the stage and the next thing you can do. Everything else is in the actions menu. ## The moves | Move | From | To | Who | | - | - | - | - | | **Request approval** | Draft | In review | Owners, admins and the AI system's owner | | **Withdraw request** | In review | Draft | The person who asked, owners and admins | | **Send back** | In review | Draft | Owners and admins, with a reason | | **Approve** | In review | Approved | Owners and admins, when the go-live checks are met | | **Go live** | Approved | Live | Owners, admins and the AI system's owner | | **Pause** | Live | Approved, paused | Owners, admins and the AI system's owner, with a reason | | **Resume** | Approved, paused | Live | Owners, admins and the AI system's owner | | **Retire** | Any other stage | Retired | Owners and admins | | **Restore** | Retired | Draft | Owners and admins | A reason or note is up to 500 characters. **Send back** and **Pause** need one of at least 10 characters: "Say why, in at least 10 characters." ## The go-live checks **Approve** shows the checks that apply at the AI system's current [rating](/ai-systems/risk-rating), each met or not. **Approve** stays off until every one is met. **Go live** and **Resume** check them again. | Check | Low and Medium | High | Very high | | - | - | - | - | | Risk profile answered | Yes | Yes | Yes | | A submitted assessment in the last | | 6 months | 3 months | | Guardrails live | | Yes | Yes | | Approved by a second person, when your organization has two or more owners and admins | | Yes | Yes | An AI system that does a practice the EU AI Act prohibits never passes: "A prohibited practice never goes live". With only one owner or admin in your organization, that person may approve their own request at High and above, and the approval says so. These are what approval needs. [Posture](/ai-systems/posture) lists everything the rating asks for, including what does not block going live. **Request approval** needs the risk profile answered first. If it is not, the dialog offers **Answer it**. You can ask for approval before the other checks are met: "You can ask now. It can be approved once these are met." ## When the rating rises after approval If someone changes the risk profile of an Approved or Live AI system and its rating goes up, it keeps its traffic and its page says "Rated higher than when it was approved. An owner or admin approves it again." Owners and admins choose **Approve again**, which checks the go-live checks at the new rating. Until then, **Go live** and **Resume** are refused: "Claims triage assistant is rated higher than when it was approved. An owner or admin approves it again first." ## Retired A retired AI system is read only: it cannot be edited, its risk profile and guardrails cannot change and no assessment can start. Its keys stay attached to it and are refused. **Restore** brings it back to **Draft**, and it needs a new approval to go live. ## What the page says | Stage | The line under the header | | - | - | | Draft | "Not live. Requests through the AI Gateway are refused." | | In review | "Waiting for approval, asked by" the person who asked. | | Approved | "Approved. Requests are refused until it goes live." | | Approved, paused | "Paused by" the person who paused it. "Requests are refused." | | Live | Nothing. | | Retired | "Retired. It is read only and its keys are refused." | The **History** tab lists every move under **Lifecycle**, newest first, with who made it, when, and the note or reason. ## Emails | Move | Sent to | | - | - | | Request approval | Owners and admins | | Approve, Approve again, Send back | The person who asked and the AI system's owner | | Go live, Resume, Pause, Retire | The AI system's owner and the organization's owners | Nobody is emailed about a move they made themselves. See [Emails](/reference/emails). ## Errors | Message | What to do | | - | - | | "Answer the risk profile of Claims triage assistant first: Edit risk profile, in its actions menu." | Answer the risk profile, then ask again. | | "Claims triage assistant cannot be approved yet: its guardrails are not live." | Close the gaps the message lists, then approve. | | "Claims triage assistant is in review, so it cannot do that now. Reload the page to see what it can do." | Someone moved it already. Reload. | | "Claims triage assistant was just changed by someone else. Reload the page to see where it is now." | Reload. | | "Your role cannot make this move. Ask an owner or admin." | Ask an owner or admin. | | "Claims triage assistant is retired. Restore it to Draft to change it." | Restore it first. | In each message Altrum uses your AI system's name. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.