# Legal and Compliance Guide

*Audience: legal, privacy, security, compliance, and audit teams. No UI steps - for those, see the [Admin Configuration Guide](/docs/risk-register/admin-configuration-guide/).*

## Why keep a risk register

Most privacy, security, and AI frameworks require an organization to identify its risks, evaluate them, decide how to treat them, and show that it keeps doing so. A register is the record of that process. The Risk Register is designed to support the risk-management requirements of frameworks such as:

| Framework | Requirement the register supports |
|---|---|
| **ISO/IEC 27001** | Information security risk assessment and treatment (clauses 6.1.2 and 6.1.3, and clause 8), including risk owners and a treatment plan |
| **ISO/IEC 27701** and **GDPR** | Assessing risk to individuals' rights and freedoms, applying appropriate technical and organizational measures (GDPR Articles 24, 25, and 32), and following up DPIAs (Article 35) |
| **EU AI Act** | A risk management system for high-risk AI systems, maintained across the system's lifecycle (Article 9) |
| **ISO/IEC 42001** and the **NIST AI RMF** | AI risk assessment, treatment, and ongoing monitoring |
| **NIST SP 800-30 / 800-53**, **NIST CSF 2.0** | Risk assessment and the Govern and Identify functions |
| **DORA**, **NIS2**, **NYDFS Part 500** | Documented ICT and cybersecurity risk management, including third-party risk |

The register supports these requirements; it doesn't by itself make you compliant with them. How you scope, score, and treat risks remains your organization's decision.

## What the register demonstrates

| You need to show | What the register gives you |
|---|---|
| Risks are identified systematically | Risks seeded from a library grounded in public sources (NIST SP 800-30, ISO/IEC 29100, the MIT AI Risk Repository, and others), from your own assessments, and from imported registers - each with its **Sources** recorded |
| Risks are evaluated consistently | Every risk is scored on the same 5 × 5 likelihood × impact scale with fixed bands, as both **inherent** and **residual** risk |
| Each risk has an accountable owner | Owners and reviewers on every risk, and a **Review status** for sign-off |
| Treatment is decided and tracked | A treatment type (mitigate, accept, transfer, avoid), treatment details, and a due date on every risk |
| Controls address the risk | Each risk links to the specific clauses of your frameworks that address it, marked as already **In place** or as **Planned treatment** |
| Controls actually operate | Linked Relyance policies test those controls against your live data, systems, and AI, and surface issues when they fail - evidence of effectiveness, not just design |
| Risks are reviewed over time | A field-level audit log of every change, with who and when, and before and after values |

## Accepted risks

Choosing **Accept** as the treatment records a decision to live with a risk. For an audit, record the rationale and any conditions in **Treatment details**, make sure an owner is assigned, and set **Review status** to **Approved** once the decision is signed off. The audit log keeps the history of who changed the treatment and when.

## AI-drafted risks

Relyance can draft risks from your documents, imported registers, and assessment answers, and suggest scores for them. Those drafts are proposals: nothing is added to the register until a person reviews and confirms it, and the person who adds it is recorded. Treat suggested scores as a starting point and rescore against your own risk criteria.

## Monitoring is evidence, not a guarantee

A risk marked **Monitored by Relyance** has at least one policy checking its controls in the systems Relyance is connected to. Issues reflect what those policies can observe in your environment. A risk with no open issues means its linked policies found nothing to raise in that scope - it isn't, on its own, proof that the risk is eliminated. Risks that can't be checked on the platform, such as contractual or organizational risks, are tracked through their controls, treatment, and reviews.

## Related

- [Audit Log and Reporting](/docs/risk-register/audit-log-and-reporting/) - what's captured and how to report on it.
- [Assessments: Legal and Compliance Guide](/docs/assessments/legal-and-compliance-guide/) - DPIAs and vendor risk assessments, which feed risks into the register.
