Independent · Vendor-neutral

The map for AI risk.

AI systems and autonomous agents are creating risks that existing controls and insurance were never designed for. The standards are multiplying and the market is moving weekly. AI Risk Atlas is the neutral reference that explains what exists, how it works, and what it means.

Start with the basics See the landscape

Start here

Follow the full path →

New to AI risk? Read these three in order, then keep going through the reference from there.

Step 1

Introducing AI Assurance and AI Insurance

A plain-language introduction to the two disciplines that manage AI's downside: assurance (proving a system behaves acceptably) and insurance (paying out when it does not), and how they connect.

Step 2

What Is AI Assurance?

AI assurance is the controls and evidence that justify trust in an AI system: evaluation, security, monitoring, governance, and independent audit. A guide to the categories, the players, and where the market is heading.

Step 3

What Is AI Insurance?

AI insurance is cover for financial loss caused by an AI system. A guide to how existing policies fall short, the silent AI problem, the new AI-specific products, and the gap they are racing to fill.

Or follow a guided path

Prefer to dive in by role? Each path is a short, ordered set of readings for a specific audience.

For business leaders

What it means for your organization

What AI assurance and AI insurance are, why both matter, and how the standards and regulations bite. No implementation detail.

For risk and security operators

Implement assurance and prepare for regulation

The AI risk stack, the standards landscape, and how the controls and risk-transfer layers fit together.

For the insurance market

Price the AI risk surface

Where existing lines are exposed, the new AI-specific products, and the assurance evidence that underwrites them.

The AI risk stack at a glance

Full breakdown →

Model riskThe AI itself: hallucination, bias, drift, prompt injection

EvalsPerformance cover

Operational riskThe AI inside a process: integration faults, scale, provider concentration

MonitoringTech E&O

Governance riskNo owner, no policy, failure to govern

ISO 42001D&O

Liability and legal riskWho is on the hook: distributed causation, regulatory exposure

ContractsAI liability cover

Reputation and trust riskBrand damage and lost trust after a visible failure

TransparencyManage, not transfer

Risk flows upward; controls are cheapest low in the stack. Each layer maps to a control and a transfer option.

Go deeper

See the full reference →

Once you have the basics, the reference goes into the AI risk stack, the standards, and how assurance and insurance fit together.

The AI Risk Stack

AI risk is not one thing. It stacks in five layers, from the model itself up to reputation. Naming each layer lets you map it to a control (assurance) and a risk-transfer option (insurance).

The Standards Landscape Explained

There is no single AI rulebook. There is a patchwork of voluntary standards you adopt and binding regulations you must obey. A guide to ISO 42001, the NIST AI RMF, the EU AI Act, and the regional rules, and how they fit together.

Latest

News & analysis →

Analysis

When Does AI Insurance Become Required?

AI insurance is optional until something makes it unavoidable. The argument that two catalysts, regulatory compulsion or a high-profile loss, decide when cover shifts from nice-to-have to table stakes.

For AI agents & builders

Read AI Risk Atlas as a machine

Structured datasets, full-corpus text, and clean per-page markdown for AI assistants and answer engines. A live MCP server is planned.

Machine access