Technical

Generative AI (IAG)

Systems generating novel content (text, images, audio, code).

How the laws define this

Both regulatory instruments agree that Generative AI (IAG) consists of systems or models trained on data to generate or create novel content, specifically highlighting formats such as text, images, audio, and code. However, they diverge significantly in their regulatory focus and context. Under Decreto 2726/2025 – Protocolo para la Adopción y Uso de Tecnologías de Inteligencia Artificial Generativa en la Administración Pública (Provincia de Santa Fe), the concept is narrowed to govern downstream deployment within public administration, framing these technologies as non-binding assistive tools subject to usage rules, risk classification, transparency obligations, and mandatory human validation. Conversely, Provvedimento n.329 (20 May 2024) - Nota informativa su web scraping per finalità di addestramento di intelligenza artificiale generativa (Garante per la protezione dei dati personali) applies the term upstream to the data collection phase, using it to frame the types of AI systems for which large-scale scraped datasets are assembled for training purposes.

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
Argentina	Decree 2726/2025 – Protocol for the Adoption and Use of Generative Artificial Intelligence Technologies in Public Administration (Province of Santa Fe)	In Force	This definition narrows generative AI to assistive tools in public administration whose non-binding outputs require human validation, plus usage rules, risk classification and transparency obligations.
Italy	Provision No. 329 (20 May 2024) - Informative Note on Web Scraping for the Purpose of Training Generative Artificial Intelligence (Data Protection Authority)	In Force	This definition frames generative AI as models trained on large-scale scraped datasets, without usage rules but in the context of data protection measures.

What it triggers

In the Province of Santa Fe public administration, it triggers treatment as assistive tools whose outputs are non-binding and require human validation, plus usage rules, risk classification and transparency obligations.source
In the Italian Informative Note, it sets out measures for website operators to protect personal data from web scraping used to train such systems.source

Definitions (2)

Generative AI (IAG) denotes AI systems whose primary function is to create novel content (text, images, audio, code) based on training data; treated by the Protocol as assistive tools whose outputs are non‑binding and require human validation. The definition narrows the category for purposes of usage rules, risk classification and transparency obligations within the public administration.

Decreto 2726/2025 – Protocolo para la Adopción y Uso de Tecnologías de Inteligencia Artificial Generativa en la Administración Pública (Provincia de Santa Fe)Definition 1 of 2

Models and systems trained on large-scale datasets to generate new content (text, images, code, audio, etc.); in the note, this term frames the types of AI systems for which scraped datasets are often assembled.

Provvedimento n.329 (20 May 2024) - Nota informativa su web scraping per finalità di addestramento di intelligenza artificiale generativa (Garante per la protezione dei dati personali)Definition 2 of 2

Not to be confused with

Generative AI (GenAI)

Generative AI (GenAI) is a plain definition of AI producing new content, while these add contexts of public administration rules or web scraping data protection.

IAGen

IAGen refers to generative systems producing novel content, similar to the core but these definitions add regulatory contexts like human oversight or training data issues.

Questions and answers

What rules apply to generative AI in Santa Fe public administration?

Under Decree 2726/2025, generative AI is treated as assistive tools with non-binding outputs requiring human validation. It includes usage rules, risk classification and transparency obligations supervised by the Secretaría de Tecnologías para la Gestión.

Does Italy's data protection authority address web scraping for generative AI training?

Provision No. 329 guides website operators on protecting personal data from automated web scraping for training generative AI models and systems. It is a non-binding informative note issued by the Garante.

How do these definitions treat generative AI outputs or training data?

The Santa Fe definition requires human validation of non-binding outputs in public administration. The Italian note focuses on protecting data from scraping for training such systems.