Human supervision
Requirement for human oversight of AI decision-support systems.
How the laws define this
Both regulatory sources establish that AI systems must serve to assist human decision-making, requiring human review and control over automated outputs to protect human agency and judgment. They agree that human personnel must maintain oversight rather than permitting unreviewed automated outcomes. However, the framework in Projeto de Lei nº 5.051/2019 (Principles for the use of Artificial Intelligence in Brazil) diverges by defining supervision as a flexible statutory requirement whose nature and intensity depend directly on the type, gravity, and implications of the supported decision, specifically incorporating tasks like monitoring, review, and the ability to intervene in outputs. In contrast, Resolución 4/2025 de la Subsecretaría de Gobierno Digital del Ministerio de Gobierno - Directrices de uso de Inteligencia Artificial Generativa en la Administración Pública de la Provincia de Buenos Aires emphasizes that human agents must maintain ultimate control over decisions or outputs, framing this mandate especially around critical public service contexts.
Synthesised from the 2 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 |
|---|---|---|---|
| Buenos Aires Province, Argentina | Resolution 4/2025 - Guidelines for the Use of Generative Artificial Intelligence in the Public Administration of the Province of Buenos Aires | In Force | This definition narrows human supervision to a core principle focused on ultimate human control and review in critical public service contexts to preserve agency and judgment. |
| Brazil | Bill No. 5.051/2019 (Principles for the use of Artificial Intelligence in Brazil) | Withdrawn | This definition widens human supervision to include monitoring, review, intervention in outputs, and supervision intensity scaled to decision type, gravity, and implications. |
What it triggers
Definitions (2)
The statutory requirement that AI systems act as auxiliary tools to human decision-making, with supervision whose nature and intensity are compatible with the type, gravity and implications of the decisions being supported (Article 4 §1); encompasses monitoring, review and the ability to intervene in automated outputs.
A core principle mandating that human agents maintain ultimate control and review over decisions or outputs generated by AI systems, especially in critical public service contexts, to preserve human agency and judgment.
Not to be confused with
Significant and competent human supervision emphasizes enabling understanding and override of AI outputs, while these definitions focus on control, review, or scaled intensity.
Meaningful human oversight stresses proportionate involvement to risk, whereas these definitions highlight ultimate control in public services or scaled supervision by decision gravity.
Human agency and oversight prioritizes preserving meaningful human decision-making and control, similar to definition 1 but without specifying public service contexts.
Questions and answers
What does human supervision require in Buenos Aires public administration?
It mandates that human agents maintain ultimate control and review over AI-generated decisions or outputs, especially in critical public service contexts. This preserves human agency and judgment.
How is the intensity of human supervision determined under Brazilian Bill No. 5.051/2019?
The supervision's nature and intensity must be compatible with the type, gravity, and implications of the decisions being supported by AI. It encompasses monitoring, review, and the ability to intervene in automated outputs.
Does human supervision apply only to public sector AI use?
Definition 1 limits it to critical public service contexts in Buenos Aires Province. Definition 2 applies more broadly to AI systems as auxiliary tools in human decision-making.
Related Terms
Significant and Competent Human Supervision
Human oversight enabling understanding and override of AI outputs....
Supervisor
Human designated to oversee an AI system and bear civil liability....
Meaningful human oversight
Appropriate human involvement proportionate to risk....
Human agency and oversight
Requirement to preserve meaningful human decision‑making and control....
Supervisión humana
Revisión y corrección humana de decisiones algorítmicas....