Article 14 is one of the most significant provisions for high-risk AI systems: it requires providers to design and develop systems so that deployers can meaningfully oversee and, if necessary, override them. The human oversight requirement reflects the EU's fundamental rights framework by ensuring that consequential decisions remain subject to human review.
Required Human Oversight Capabilities
Article 14(4) specifies the minimum capabilities that high-risk AI systems must enable for human overseers: the ability to fully understand the system's capacities and limitations; the ability to monitor the operation for signs of anomalies; the ability to interpret outputs correctly given the context; the ability to disregard, override, or reverse a system output; and the ability to intervene or halt the system through a stop button or similar procedure. These capabilities must be built into the system design -- they cannot be purely procedural or organisational measures.
Regulation (EU) 2024/1689 — Article 14(4)(e)
“Be able to decide, in any particular situation, not to use the high-risk AI system or otherwise disregard, override or reverse the output of the high-risk AI system.”
- Human oversight is a technical design requirement, not just a policy or procedure
- Deployers must be trained to exercise oversight effectively -- this links to the Article 4 literacy obligation
- Article 14(5) addresses "automation bias" -- systems must be designed to counter over-reliance by human overseers
- High-risk systems used in law enforcement have additional oversight requirements under Article 14(3)
- The duty to override is a right of the deployer, not merely a last-resort emergency procedure