Lead efforts to establish shared criteri…

BackCoordinate evaluations and safety testing

This page is still being polished. If you have thoughts, please share them via the feedback form.

Data on this page is preliminary and may change. Please do not share or cite these figures publicly.

Coordinate evaluations and safety testing

Somai (2025)|LLM classified

Mitigation Taxonomy

2Organisation

2.2Risk & Assurance

2.2.2Testing & Evaluation

Red teaming, capability evaluations, adversarial testing, and performance verification.

Also in Risk & Assurance

2.2.1 Risk Assessment2.2.3 Auditing & Compliance2.2.4 Assurance Documentation

Additional Informationp. 4

Stage: Detection; Stakeholder: National Government: AISI; Additional information: AI developers and researchers should refine detection by developing standardised benchmarks and improving their reliability and validity. Developers should enhance detection of control-undermining capabilities.

LLM Classification Details

Reasoning

National government coordinates safety testing and evaluations across organizations to assess AI system capabilities and risks.

Code: 2.2.2Version: v0.6Classified: Jan 27, 2026

Other mitigations from Somai (2025) (38)

Monitor critical capability levels

2.2.2 Testing & Evaluation

Lifecycle:Other (multiple stages)Actor:DeveloperAIRM:Measure

Identify early warning signs and emergent capabilities

2.2.1 Risk Assessment

Lifecycle:Build and Use ModelActor:DeveloperAIRM:Measure

Establish standardised benchmarks and reporting

3.2.1 Benchmarks & Evaluation

Lifecycle:Verify and ValidateActor:DeveloperAIRM:Measure

Implement compute monitoring and anomaly detection

1.2.3 Monitoring & Detection

Lifecycle:Other (multiple stages)Actor:Infrastructure ProviderAIRM:Measure

Enhance hardware and supply chain oversight

2.3.3 Monitoring & Logging

Lifecycle:Other (stage not listed)Actor:Infrastructure ProviderAIRM:Govern

Lead efforts to establish shared criteria for AI LOC

3.2.2 Technical Standards

Lifecycle:Plan and DesignActor:Governance ActorAIRM:Govern

View all 38 mitigations from this source →

Source Document

Strengthening Emergency Preparedness and Response for AI Loss of Control Incidents

Somani, Elika; Friedman, Anjay; Wu, Henry; Lu, Marianne; Byrd, Christopher; van Soest, Henri; Zakaria, Sana (2025)

As artificial intelligence (AI) systems become increasingly embedded in essential infrastructure and services, the risks associated with unintended failures rise. Developing comprehensive emergency response protocols could help mitigate these significant risks. This report focuses on understanding and addressing AI loss of control (LOC) scenarios where human oversight fails to adequately constrain an autonomous, general-purpose AI.

View source DOI: 10.7249/RRA3847-1

Classification

AI Lifecycle Stage

Verify and Validate

Testing, evaluating, auditing, and red-teaming the AI system

Responsible Actor

Governance Actor

Regulator, standards body, or oversight entity shaping AI policy

NIST AI RMF Function

Govern

Policies, processes, and accountability structures for AI risk management

Measure