Advanced AI systems
Models or systems with high‑impact outputs or autonomy.
How the laws define this
Regulatory guidance and international frameworks agree that advanced AI systems are characterized by their significant capabilities and their potential to create substantial impacts, risks, or harms. These definitions consistently associate advanced systems with features such as autonomy, emergent or unpredictable behaviors, or classification as foundation models and generative AI. The principal divergence lies in the administrative purpose assigned to the term. The Buku Panduan Tata Kelola Kecerdasan Artifisial Perbankan Indonesia (OJK Guidance on AI Governance for Indonesian Banking) uses the designation to require enhanced governance, validation, and oversight for models carrying systemic or customer impact. Conversely, the Canadian Artificial Intelligence Safety Institute (CAISI) applies domain-specific definitions to prioritize safety research and testing on systems capable of technical or societal harm. Additionally, the Hiroshima Process International Guiding Principles and its monitoring reporting framework restrict the term to sophisticated systems developed specifically by organizations subject to the Hiroshima Process International Code of Conduct.
Synthesised from the 4 statutory definitions below. Each one is quoted in full, with its source.
How jurisdictions differ
What is distinctive in each definition, binding law in force first.
| Jurisdiction | Instrument | Status | What is distinctive |
|---|---|---|---|
| Canada | Canadian Artificial Intelligence Safety Institute (CAISI) | In Force | Adds prioritization of research on systems whose failures or behaviors could cause significant societal or technical harms, applying domain-specific definitions for project selection and testing. |
| Indonesia | OJK Guidance on AI Governance for Indonesian Banking | In Force | Characterizes systems by large scale or emergent capabilities leading to significant autonomy, complexity, unpredictable behaviors, and higher systemic or customer impact in banking. |
| OECD | Reporting framework for monitoring the application of the Hiroshima Process International Code of Conduct for Organizations Developing Advanced AI Systems | In Force | Limits to AI systems with significant capabilities, often generative, developed by organizations subject to the Hiroshima Process International Code of Conduct. |
| G7 | Hiroshima Process International Guiding Principles for Organizations Developing Advanced AI Systems and International Code of Conduct for Organizations Developing Advanced AI Systems | Adopted | Specifies the most sophisticated systems, including foundation models and generative AI, that present significant capabilities and potential risks. |
What it triggers
- CAISI applies domain-specific definitions for project selection and testing to prioritize research on advanced AI systems.source
- OJK Guidance recommends enhanced governance, validation, and oversight measures for advanced AI systems in Indonesian banking.source
- OECD Reporting framework enables organizations developing advanced AI systems to voluntarily report adherence to the Hiroshima Process International Code of Conduct.
- G7 Hiroshima Process establishes safety principles and a voluntary code of conduct for organizations developing advanced AI systems.source
Definitions (4)
Advanced AI systems generally denote models or systems with capabilities that can produce high‑impact outputs or act with a degree of autonomy; CAISI applies domain‑specific definitions for project selection and testing. The term is used to prioritize research on systems whose failures or behaviours could cause significant societal or technical harms.
Models characterized by large scale or emergent capabilities that confer significant autonomy, complexity, or unpredictable behaviors; such systems may have higher systemic or customer impact and therefore require enhanced governance, validation and oversight measures.
The most sophisticated artificial intelligence systems, including foundation models and generative AI systems, that present significant capabilities and potential risks.
AI systems with significant capabilities, often generative, developed by organizations that are subject to the Hiroshima Process International Code of Conduct.
Not to be confused with
High-impact/high-risk system refers to systems likely to significantly affect safety, rights, or critical infrastructure, while advanced AI systems focus on capabilities like high-impact outputs or autonomy.
High-impact artificial intelligence system emphasizes potential for serious harms or discriminatory outcomes, whereas advanced AI systems highlight capabilities, autonomy, or sophistication.
Questions and answers
What makes an AI system advanced under CAISI?
CAISI denotes advanced AI systems as those with capabilities producing high-impact outputs or acting autonomously. It applies domain-specific definitions for project selection and testing. The focus is prioritizing research on systems risking significant societal or technical harms.
How does OJK define advanced AI in banking?
OJK characterizes advanced AI as models with large scale or emergent capabilities conferring significant autonomy, complexity, or unpredictable behaviors. Such systems may have higher systemic or customer impact. Enhanced governance, validation, and oversight are recommended.
What are advanced AI systems in the G7 Hiroshima Process?
The G7 defines them as the most sophisticated AI systems, including foundation models and generative AI, with significant capabilities and potential risks. It establishes safety principles and a voluntary code of conduct for developers.
Related Terms
AI systems
Systems using automated models to make/assist decisions or generate content....
AI System
Machine-based system that infers outputs from inputs....
High-impact/high-risk system
Systems likely to significantly affect safety, rights, or critical infrastructure....
High-Impact AI Systems
AI systems whose use poses significant risks to rights or safety....
High-impact artificial intelligence system
AI systems that may cause serious harms or discriminatory outcomes....