High‑risk AI

A principles‑based category focused on AI systems whose use, given context and deployment, could result in significant severity or scale of harm (including impacts to safety, health, fundamental rights, or critical services). The designation is determined by assessing potential harms across the AI lifecycle and supply chain, not solely by technical characteristics.

A working classification adopted by the amendment that follows the approach of the EU AI Act to denote AI systems whose deployment presents substantial safety risks or significant implications for fundamental rights, and which therefore require heightened oversight, conformity assessment and mitigation measures under applicable law.

A category of AI systems referenced in alignment with the EU AI Act, encompassing systems whose use may substantially affect health, safety, legal status or fundamental rights and which are therefore subject to stricter regulatory obligations and oversight.

Systems classified by the draft as posing significant potential impact on health, safety, fundamental rights, or critical infrastructure; such systems are subject to enhanced obligations including pre‑deployment conformity assessments, detailed technical documentation, logging, registration, mandatory Fundamental Rights Impact Assessments, and post‑market monitoring.

AI systems deployed in sectors or applications that may substantially affect life, safety, fundamental rights, public order, or other high‑impact outcomes (e.g., healthcare, finance, employment, critical infrastructure); designation triggers heightened obligations such as risk management, prior notice, documentation, and conformity assessment.

A legal category covering AI applications and deployments that impose a significant risk to life, bodily integrity, property, or fundamental rights; examples cited include biometric identification, critical infrastructure control, personnel assessment, emergency services, credit or essential public services, and state investigative/immigration uses. Systems classified as high‑risk are subject to enhanced pre‑deployment assessments, mandatory technical and organizational safeguards, ongoing monitoring, documentation, and stricter transparency and conformity obligations.

AI systems classified as likely to materially affect fundamental rights, safety, health, or critical interests and therefore subject to enhanced obligations (risk frameworks, documentation, pre‑deployment testing, registration, and conformity assessment). Sectoral regulators will define specific high‑risk categories in subordinate instruments.

AI systems that, by virtue of their purpose or application, can have significant impacts on safety or fundamental rights; classification is informed by the EU AI Act and guides prioritisation of standards, conformity assessment and regulatory controls. The Roadmap uses this category to focus testing, certification and liability mitigation measures for systems with elevated risk.

AI systems characterised by their potential impact on fundamental rights, public safety, essential services, or national critical infrastructure; such classification triggers enhanced compliance obligations under the strategy (e.g., registration, mandatory algorithmic impact assessments, audits, and remediation).

Deployments of AI that materially affect safety, fundamental rights, or essential services and therefore require enhanced oversight, mandatory impact assessment, and risk mitigation measures under CERAI guidance.

AI systems that materially affect life, safety, health, or basic rights and that operate in critical sectors (e.g., energy, water, medical care and devices, critical infrastructure, employment/credit decisions, transport, and public sector systems), subject to mandatory impact assessments, prior verification, safety testing, and stricter transparency and mitigation requirements.

Classes of AI systems identified as posing significant risks to safety, fundamental rights, access to essential services, or other critical public interests; such systems are subject to mandatory registration, enhanced risk‑management, conformity assessment, and ongoing supervisory obligations.

AI systems whose malfunction, biased operation, or erroneous outputs could cause significant physical, psychological, economic, or societal harm to individuals or groups; such systems are subject to enhanced assessment, monitoring, mitigation, and potential regulatory oversight under the notification.