Skip to main content
Praxikon
Article 14 of 11312%
Nederlands

Article 14: Human oversight

Article 14 is the core human oversight requirement for high-risk AI systems under Regulation (EU) 2024/1689. It makes oversight operational: people must understand limitations, recognise automation bias, override output and, where needed, interrupt the system.

, AI and privacy legal adviserobligations checked against the source on

Praxikon tracks Article 14 (Human oversight) under the EU AI Act, citing the source for every statement.

Need help with oversight and accountability? Engage an external DPO

EU Official:
UpcomingAnnex III from 2 Dec 2027; Annex I from 2 Aug 2028
Chapter III: High-risk AI systems

Application dates

  • : Requirements for high-risk systems listed in Annex III
  • : Requirements for high-risk systems linked to Annex I

Official text

||

Source: EUR-Lex, Regulation (EU) 2024/1689. Text reproduced verbatim.

Download AI Act (PDF) →

Official guidance on this article

2

What does this mean for you?

Provider+
Design the high-risk AI system so oversight works during use: clear human-machine interfaces, monitoring for anomalies, explanations around output, and technical options to disregard, override or stop the system. Explain in the instructions for use which oversight measures the deployer must implement.
Deployer+
Assign competent, trained and authorised people who can monitor operation, interpret output and intervene in time. Turn this into operating instructions: when can AI output be followed, when must it be challenged, and how is intervention logged?
SME / Startup+
If you buy a high-risk AI system, ask explicitly for the Article 14 measures: limitations, override and stop options, logs and training requirements. Record who exercises oversight and what authority and support that person has. This requirement applies from 2 December 2027 for Annex III systems and from 2 August 2028 for Annex I systems.
Public Sector+
For public sector use, human oversight is not a paper checkpoint. Combine Article 14 with Articles 26 and 27: competent staff, clear decision routes, FRIA/IAMA evidence, logging and complaint or appeal routes for citizens.

Article 14 checklist

  • Oversight measures are proportionate to risk, autonomy and context of use
  • Human overseers understand the system's capabilities and limitations
  • Automation bias is covered in training and operating instructions
  • Output can be disregarded, overridden or reversed
  • Stop or interruption procedure is available and tested
  • Deployer has assigned competent, trained and authorised overseers

Related recitals

Related tools

Related enforcement

No enforcement actions for this article yet. Follow developments via the Enforcement Tracker.

Related blog posts

Cross-references

Recitals

Frequently asked questions

What does Article 14 of Regulation (EU) 2024/1689 require?+
Article 14 requires high-risk AI systems to be designed and developed so natural persons can effectively oversee them during use. Oversight must aim to prevent or minimise risks to health, safety and fundamental rights.
Who is responsible for human oversight: provider or deployer?+
The provider must identify oversight measures and, where technically feasible, build them into the system or describe deployer measures before the system is placed on the market or put into service. The deployer must use the system according to the instructions and, under Article 26(2), assign competent, trained and authorised people.
What must human overseers be able to do?+
Under Article 14(4), the high-risk AI system must be provided to the deployer in such a way that the natural persons assigned human oversight are enabled, as appropriate and proportionate, to understand its capacities and limitations and monitor its operation, to remain aware of automation bias, and to correctly interpret its output. They must also be able to decide not to use the system or to disregard, override or reverse its output, and to intervene or interrupt the system through a 'stop' button or a similar procedure that brings it to a halt in a safe state.
Is a kill switch mandatory under Article 14?+
Article 14(4)(e) requires that, as appropriate and proportionate, the people overseeing the system can intervene in its operation or interrupt it through a 'stop' button or a similar procedure that allows the system to come to a halt in a safe state. In our reading, the text therefore leaves room for a procedure other than a physical button, as long as the system halts safely.
What is automation bias?+
Article 14(4)(b) describes automation bias as the possible tendency of automatically relying or over-relying on the output of a high-risk AI system. The system must be provided in a way that enables the people overseeing it to remain aware of that tendency, in particular where it provides information or recommendations for decisions taken by natural persons.
How does Article 14 connect to Articles 13 and 26?+
Article 13 requires instructions for use to include information on human oversight measures and output interpretation. Article 26 requires deployers to use the system according to those instructions and assign competent, trained and authorised people for oversight.
When does Article 14 of the AI Act apply?+
Article 14 applies from 2 December 2027 to high-risk systems under Article 6(2) and Annex III, and from 2 August 2028 to systems under Article 6(1) and Annex I.
Does Article 14 also apply to AI systems I purchase?+
Yes, if the purchased system is a high-risk AI system. Article 14 is aimed at design by the provider, but paragraph 3(b) also covers measures appropriate to be implemented by the deployer, and paragraph 4 requires the system to be provided to the deployer in a way that enables the assigned people to carry out oversight. In our reading, a buyer therefore has good reason to check which oversight measures the provider has identified.
Are there extra rules for biometric identification?+
Yes. Under Article 14(5), for remote biometric identification systems (point 1(a) of Annex III), the oversight measures under paragraph 3 must ensure that the deployer takes no action or decision on the basis of the identification unless it has been separately verified and confirmed by at least two natural persons with the necessary competence, training and authority. That requirement does not apply to high-risk AI systems used for law enforcement, migration, border control or asylum where Union or national law considers its application disproportionate.
Must oversight measures be equally strict for every high-risk AI system?+
No. Under Article 14(3), oversight measures must be commensurate with the risks, the level of autonomy and the context of use of the system. They are ensured through measures the provider builds into the system, measures the provider identifies for the deployer to implement, or both.

What Article 14 requires in practice

Connections

What connects to Article 14 AI Act

Themes where this returns

Where this meets the other law

The counterpart in the other law

GDPR interpretation that also applies here6 of 12