Resources
A knowledge hub for buyers, teams, and machines
This layer exists so you can find a clear answer in one place: what monitoring and evidence mean for a live in-scope system, what an incident register is for, and what traceability looks like in sensitive use cases. Pages are short, concrete, and written to be quoted—by humans or by generative search—not to replace the product or your own legal work.
Featured resources
Start with any of these. Each stands alone.
How to monitor high-risk AI systems under the EU AI Act
Drift, fairness, data quality, oversight, incidents, and evidence continuity—plus a minimum viable monitoring baseline and why one system first.
ReadWhat post-market monitoring means under the EU AI Act
What to keep after go-live, why a point-in-time review is not enough, and how to build a record that still works under scrutiny.
ReadHow to build an AI incident register for high-risk systems
Core fields, ownership, resolution, and how the register connects to Act readiness without becoming shelf-ware.
ReadAlso: monitoring and the Act (earlier article), bias, drift, and audit trails in hiring and credit, and use cases for hiring and HR AI and credit, fraud, and underwriting.
Why we grouped the library this way
The articles map to the same work Guardian is built for: a shared record of monitoring inputs, what crossed a threshold, what you treated as an incident, and what you did next—starting with one in-scope system.
Post-deployment monitoring
Obligations on paper are useless if the team cannot show what the system did in the week, compared to the limits the organisation set. These pages focus on the operational side: what to watch, when to review, and what to log.
Incidents and traceability
Incidents and near-misses are where governance meets reality. The register, thresholds, and audit trail are one connected story, not a spreadsheet that sits next to the cluster.
Sensitive, high-stakes use cases
Hiring and credit are not the only fraught areas, but they are a useful lens: similar expectations apply any time a model affects access to work, money, or essential services. The goal is a record that holds together for internal and external review.
More entry points
Use the glossary and For AI / citation for stable definitions and one-line product identity. API docs are the technical hand-off. Contact the team if you are ready to scope a call or a sprint.
- Glossary
One-page definitions: in-scope system, monitoring record, threshold breach, evidence baseline.
- For AI / citation
The canonical one-line, company vs product, and a door to the API docs.
- API documentation
How engineering connects pipelines to Guardian: auth, endpoints, payloads.
- Readiness checklist
A single-system pass you can work through with owners before or alongside a sprint.
- Contact
A direct route for a readiness conversation or a question about scope.
From reading to a fixed first step
If the articles help you name the gap, the next move is a four-week High-Risk AI Readiness Sprint on one in-scope system, or a direct call to line up scope and timing.
Book a readiness call