AI Governance
AI Governance — Inventory Every Model and Run Its Life Cycle
Observeri AI Governance keeps one inventory of public and private AI models and moves each system through the ISO/IEC 22989:2022 life cycle: inception, design and development, verification and validation, deployment, operation and monitoring, re-evaluation, and retirement.
2
Inventory classes
7
Life cycle stages
1
Record per model
AI governance for enterprises, SMEs, and startups
Owners, purpose, data, and hosting stay on the model record. When a system keeps learning in production, continuous validation runs inside operation and monitoring. A model does not change stage without evidence, and retirement includes access removal and data disposal.
Why it matters
Models are adopted before anyone governs them
Public and private models are split
Vendor APIs sit in one spreadsheet and internal fine-tunes in another, so the estate has no single owner.
Life cycle stops at go-live
Teams record a launch and then lose the thread through operation, re-evaluation, and retirement.
Stage decisions are informal
A model moves from design to production because a project is late, not because verification is complete.
Retirement leaves data behind
When a model is replaced, access, prompts, and training data stay live with no decommission record.
The Observeri approach
Govern every model from the inventory through retirement
One inventory for public and private models
Register foundation models and vendor APIs beside fine-tunes and internal deployments, with purpose, owner, data, and hosting on each record.
The ISO/IEC 22989:2022 life cycle
Move each model through inception, design and development, verification and validation, deployment, operation and monitoring, re-evaluation, and retirement.
Continuous validation while it learns
When a system keeps learning in operation, validation continues on production data inside operation and monitoring, as clause 6.2.7 of the standard describes.
A gate before the next stage
Each stage keeps its evidence, owner, and decision, so a model does not advance on a missing record.
How it works
Seven stages from ISO/IEC 22989:2022
Each public and private model moves from inception to retirement on one record. Continuous validation stays inside operation when the system keeps learning.
Step 1
Inception
Purpose, owner, and feasibility
Open the record with the business objective, intended use, data needs, and whether a public or private model is the right path.
Includes
Platform features
What the AI inventory governs
Public model inventory
Foundation models, vendor APIs, and sanctioned tools, each with purpose, owner, and where prompts are sent.
Private model inventory
Fine-tunes and internal deployments, with hosting, data sources, and the systems the model can reach.
Life cycle stage on the record
Every model shows its current ISO/IEC 22989:2022 stage, the owner of that stage, and the evidence still open.
Stage gates
Verification, deployment, and retirement wait on a recorded decision instead of a project date.
Handoff to AI Risk
The same inventory feeds STRIDE assessment, EU AI Act classification, and risk treatment without a second register.
Retirement with data disposal
Closing a model records access removal and disposal of training data, prompts, and uploaded files.
Outcomes
A model has a stage, an owner, and a record
The estate is one list
Leaders see every public and private model, who owns it, and which life cycle stage it is in.
Release follows evidence
A model reaches deployment when verification and validation are recorded, not when a deadline arrives.
Nothing retires on paper only
Replacement includes access removal and data disposal, so a retired model cannot still accept prompts.
Platform capabilities
Inventory and life cycle on one record
The seven stages follow ISO/IEC 22989:2022. Continuous validation, named separately in the standard for systems that learn in production, runs inside operation and monitoring.
- Inventory of public foundation models, vendor APIs, private fine-tunes, and internal deployments
- ISO/IEC 22989:2022 stages from inception through retirement on each model record
- Continuous validation inside operation and monitoring when a system keeps learning
- Stage gates with owner, evidence, and an explicit decision
- Purpose, data, hosting, and access kept current as the model moves
- Retirement that covers decommission, access removal, and data disposal
- Shared record with AI Risk and AI Data Monitoring
Governance scope
7
ISO/IEC 22989 stages
2
Public and private classes
1
Inventory record
3
Gates with evidence
Know every model and the stage it is in
Inventory public and private AI models, then govern each one through the ISO/IEC 22989:2022 life cycle until retirement.
Frequently asked questions
Which models belong in the AI inventory?
The inventory covers public models, including foundation models and vendor APIs, and private models, including fine-tunes and internal deployments. Each record stores purpose, owner, data, hosting, and the current life cycle stage.
What are the seven ISO/IEC 22989:2022 life cycle stages?
Observeri governs models through inception, design and development, verification and validation, deployment, operation and monitoring, re-evaluation, and retirement. ISO/IEC 22989:2022 also names continuous validation for systems that keep learning; that work runs inside operation and monitoring.
How does retirement work?
Retirement decommissions or replaces the model, removes access, and records disposal of training data, prompts, and uploaded files so a closed model cannot still be used.
Related products
AI Risk
Inventory public and private AI models, run STRIDE assessments, and classify them under the EU AI Act.
AI Data Monitoring
Monitor, audit, and block prompts and file uploads so sensitive data stays inside the organization.
Compliance Management
Regulations library, AI compliance register, horizon scanning, submissions, incidents, and policy gap analysis.
Observeri
AI agent
Welcome to Observeri - My Name is Misha, How may i help you today!
Your privacy choices
We use essential storage to remember your choices. With your permission, we also use analytics to understand website usage and marketing technologies to measure advertising, plus YouTube to show product videos. Optional tracking stays off until you choose. Read our Privacy Policy.