Systemic Risk
Risk that a general-purpose AI model could cause widespread, significant harm to public health, safety, security, or society.
How the laws define this
Regulatory authorities agree that systemic risk involves broad, widespread harms or negative consequences that cascade across sectors, extend beyond a single domain, and impact fundamental rights, public safety, democratic processes, or critical and national infrastructure. The main divergence lies in whether systemic risk is framed as a broad governance challenge or bound to specific AI model thresholds. The Central AI Risk Function (CAIRF) defines it through wide-ranging impacts that potentially necessitate whole-of-government responses and cross-sector mitigation strategies. In contrast, the Artificial Intelligence Act (Regulation (EU) 2024/1689) and the General-Purpose AI Code of Practice tie systemic risk directly to general-purpose AI models based on their scale, deployment, or capabilities. Furthermore, the Artificial Intelligence Act and General-Purpose AI Code of Practice establish quantitative criteria—specifically cumulative training compute exceeding 10^25 FLOPs—to identify high-impact capabilities and trigger enhanced regulatory obligations.
Synthesised from the 3 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. Showing 2 of the 3 instruments that define this term; all are listed under Definitions below.
| Jurisdiction | Instrument | Status | What is distinctive |
|---|---|---|---|
| European Union | General-Purpose AI Code of Practice (voluntary Code of Practice for GPAI providers) | In Force | Focuses on general-purpose AI models identified by training compute thresholds like exceeding 10^25 FLOPs, unlike the broader cross-sectoral focus of others. |
| United Kingdom | Central AI Risk Function (CAIRF) | In Force | Emphasizes cascading risks across sectors or national infrastructure, without referencing specific AI model types or compute thresholds. |
What it triggers
Definitions (3)
A level of risk arising from an AI model or family of models whose scale, deployment or capabilities could cause broad, cross-sectoral harms to fundamental rights, public safety, democratic processes or critical infrastructure; the Regulation provides criteria and procedures to identify systemic-risk models (notably among general-purpose AI) to trigger enhanced obligations and Commission powers.
Risks that can cascade across sectors or national infrastructure, producing wide-ranging impacts beyond a single domain and potentially necessitating whole-of-government responses and cross-sector mitigation strategies.
Applies to GPAI models that are highly capable or widely used and could potentially lead to significant, widespread negative consequences. The AI Act uses criteria such as the cumulative amount of computing power used to train a model (e.g., exceeding 10^25 FLOPs) to identify models presumed to have high-impact capabilities.
Not to be confused with
Frontier AI refers to highly capable cutting-edge models that may create systemic risks, while systemic risk describes the actual level or potential of those harms.
General-Purpose AI Model with Systemic Risk is a specific legal designation under the EU AI Act, whereas systemic risk is the broader risk level triggering such designation.
Questions and answers
What triggers systemic risk designation for GPAI models in the EU?
Models are presumed systemic risk based on criteria like exceeding 10^25 FLOPs of training compute power. This applies to highly capable or widely used GPAI models with potential for significant widespread harm.
How does the UK address systemic AI risks?
The Central AI Risk Function monitors cross-sectoral risks that cascade across sectors or infrastructure. It assists regulators and departments with mitigation strategies.
What obligations apply to systemic risk AI models?
Under the EU AI Act, systemic-risk models trigger enhanced obligations and powers for the Commission. The General-Purpose AI Code of Practice provides voluntary guidelines for compliance.
Related Terms
General-Purpose AI Model with Systemic Risk
A general-purpose AI model legally designated as carrying risks that can affect the EU market or society at scale, subject to heightened obligations under the EU AI Act....
Frontier AI
Highly capable, cutting‑edge AI models (a subset of general‑purpose/foundation models) whose scale or capabilities may create systemic or catastrophic risks....
Dual-Use AI
AI systems or models that can be applied for both beneficial civilian purposes and potentially harmful applications....
AI Safety
Measures and engineering practices ensuring AI systems do not cause unreasonable harm to people, property, or the environment during intended use or foreseeable misuse....