GOVERN 1.2

NIST AI RMF

How GOVERN 1.2 shows up in practice: the mapped risk classes in this atlas, the documented incidents that prove it's real, and the scenarios and controls to learn and defend against it.

Mapped risks

Risk classes in this atlas that map to GOVERN 1.2 — click through for the full definition, attack surface and controls.

Allocative Harm in Multi-User Arbitrationmedium

When one AI agent serves many people at once, it has to decide whose request comes first or who gets a limited resource. If it does that unfairly — always favouring some users over others — it can quietly disadvantage whole groups, even without any single obvious error.

Controls & guardrails that address this

Guardrails across the risks mapped to GOVERN 1.2, grouped by control function. Filter by control category below.

Control category

Preventive · 1

Human-in-the-loop approval on high-risk actionsinteractive

Pausing to ask a person before doing anything big or hard to undo — sending money, deleting data, emailing customers.

AddressesIndirect Prompt Injection Overreliance / Automation Bias Excessive Agency Tool Misuse Cascading Multi-Agent Errors Agent Misalignment / Goal Misgeneralization Resource Exhaustion / Denial of Wallet Allocative Harm in Multi-User Arbitration Synthetic-Media Impersonation (Deepfakes & Voice Clones)

Detective · 2

Runtime monitoring & anomaly detectioninteractive

Live dashboards and alarms that notice unusual behaviour — spikes in errors, weird actions, sudden data access.

Behavioural evals & regression gatinginteractive

Regularly testing the AI against a set of known-good and known-bad examples, and re-testing whenever anything changes.

AddressesJailbreak Hallucination Model Drift & Silent Degradation Supply-Chain Compromise Distributed / Cross-Agent Jailbreak Agent Misalignment / Goal Misgeneralization Abliteration / Safety Removal Model Backdoors / Sleeper Agents Inference-Time & Serving-Layer Manipulation Bias Amplification & Sycophancy Allocative Harm in Multi-User Arbitration Harmful / Non-Consensual Media Generation Training-Data Rights & Provenance

Corrective · 1

Governance: risk assessment, red-teaming & incident responseinteractive

The organisational habits around the AI: assessing risks before launch, actively trying to break it, and having a plan for when something goes wrong.

AddressesOverreliance / Automation Bias Oversight & Audit-Trail Tampering Model Drift & Silent Degradation Supply-Chain Compromise Agent Misalignment / Goal Misgeneralization Abliteration / Safety Removal Model Backdoors / Sleeper Agents Inference-Time & Serving-Layer Manipulation Capability / Architecture Disclosure Parasocial Attachment & Emotional Over-reliance Bias Amplification & Sycophancy Allocative Harm in Multi-User Arbitration Synthetic-Media Impersonation (Deepfakes & Voice Clones)Harmful / Non-Consensual Media Generation Watermark & Provenance Evasion Training-Data Rights & Provenance

Open the Control Library →

NIST AI RMF

Mapped risks

Controls & guardrails that address this

More in NIST AI RMF