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.
Runtime monitoring, observability, performance tracking, and anomaly detection in production.
Also in Operations & Security
Stage: Detection; Stakeholder: Compute Providers; Additional information: Compute providers, national security agencies, and cybersecurity professionals could be trained to recognise LOC indicators and monitor developments in AI capabilities. Cloud providers could incorporate real-time compute monitoring and verification to flag high-risk users. Enhancing the information flow between AI developers, compute providers and governments on AI R&D would also improve detection. Governments should consider requiring developers to track and report key metrics, such as compute usage for AI R&D, as well as to disclose extreme capabilities to AISIs (Mikton 2024).
Reasoning
Enhance operational monitoring and observability of hardware and supply chain activities for abuse detection.
Monitor critical capability levels
2.2.2 Testing & EvaluationIdentify early warning signs and emergent capabilities
2.2.1 Risk AssessmentEstablish standardised benchmarks and reporting
3.2.1 Benchmarks & EvaluationImplement compute monitoring and anomaly detection
1.2.3 Monitoring & DetectionLead efforts to establish shared criteria for AI LOC
3.2.2 Technical StandardsCoordinate evaluations and safety testing
2.2.2 Testing & EvaluationStrengthening 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.
Other (stage not listed)
Applies to a lifecycle stage not captured by the standard categories
Infrastructure Provider
Entity providing compute, platforms, or tooling for AI systems
Govern
Policies, processes, and accountability structures for AI risk management
