AI Regulation Tracker / Standards and frameworks

ISO/IEC 12792: Voluntary AI Transparency Taxonomy Standard

According to the ISO catalogue entry, in November 2025 ISO and IEC published the first edition of ISO/IEC 12792, a taxonomy of transparency information elements for AI systems. It gives stakeholders a shared way to describe what information about an AI system should be disclosed, to whom, and for what purpose. It is a voluntary international standard, not a law, and it does not by itself require anyone to disclose anything.

By Anthony Guerriero · The Leveraged Years Regulatory Research · Published November 2025 · Last updated November 2025 · Standards briefing

The short version

ISO/IEC 12792:2025 is a new international standard, published in November 2025 as a first edition, that lays out a taxonomy of the information elements involved in AI transparency. In plain terms, it names and organizes the kinds of things you might tell people about an AI system, and it explains the meaning of each element and why it matters to different stakeholders. This is a voluntary standard. Buying and following it is a choice, and the document does not carry any legal force on its own. Who it affects: teams building AI governance programs, vendors writing model and system documentation, and anyone trying to line up their disclosures with management-system standards or emerging law. My read is that the value here is a shared dictionary. When a customer, an auditor, and a regulator all use different words for the same disclosure, transparency becomes an argument about vocabulary. A common taxonomy narrows that gap. It does not tell you what the law requires. The primary source is the ISO catalogue entry for the standard.

Regulatory Briefing

Instrument: ISO/IEC 12792:2025, "Information technology, Artificial intelligence (AI), Transparency taxonomy of AI systems," Edition 1
Authority: International Organization for Standardization (ISO) and International Electrotechnical Commission (IEC), joint technical committee for AI
Jurisdiction: International, voluntary. Applies wherever an organization chooses to adopt it
Status: Published, Edition 1, 2025-11
Bindingness: Non-Binding Framework. Signal, not action. A voluntary consensus standard with no legal force of its own. It becomes an obligation only if a contract, a certification scheme, or a law incorporates it by reference
Relevant date: Published November 2025 (2025-11)
Primary Source: ISO catalogue, iso.org/standard/84111.html
Workflow Impact: Gives documentation, procurement, and audit teams a common set of terms for AI transparency disclosures. It standardizes vocabulary, not the underlying legal duty to disclose
Operational Next Step: If you maintain AI system documentation, review the taxonomy against your current disclosure templates to see where your terms and its terms diverge
Last Verified: July 23, 2026

Here is what the standard actually is. Per the ISO catalogue, ISO/IEC 12792:2025 "specifies a taxonomy of information elements to assist AI stakeholders with identifying and addressing the needs for transparency of AI systems." It "describes the semantics of the information elements and their relevance to the various objectives of different stakeholders." That is the whole scope in the publisher's own words. It is a classification of the pieces of information that make an AI system transparent, plus an explanation of what each piece means and who cares about it.

What it is not: a rule. ISO and IEC publish consensus standards that organizations adopt voluntarily. Nothing in the catalogue entry says a company must follow the taxonomy, and no regulator is enforcing it. You buy the document, and you use it because it is useful, not because you are compelled to.

Why a taxonomy, and not a rulebook?

Transparency is one of those words everyone agrees on and no two teams define the same way. One group means a model card. Another means a data sheet. A third means user-facing notice that an answer came from a machine. A taxonomy does the unglamorous work of naming these things consistently so a conversation about disclosure can start from shared terms rather than crossed wires.

The standard groups the information elements and, importantly, ties them to stakeholder objectives. A developer, a deployer, an end user, and an oversight body do not want the same disclosures. Sorting the elements by who needs what, and why, is more useful than a flat list. That is the practical contribution here.

Where this fits for a US professional

This is my interpretation, not a claim from the ISO text. Voluntary AI standards rarely stay purely voluntary. They tend to get pulled into contracts, certification schemes, and, eventually, the way regulators and courts describe reasonable practice. A shared transparency vocabulary is exactly the kind of building block that management-system standards and disclosure laws lean on, because it saves everyone from reinventing the same definitions.

So the reason to look at ISO/IEC 12792 now is not that it binds you. It does not. The reason is that if your documentation already speaks a compatible language, you are better positioned when a customer's procurement team, an auditor, or a future obligation asks you to show your transparency practices in an organized way. Aligning vocabulary early is cheap. Retrofitting it after a deal or an audit demands it is not.

What to do with it

If you own AI system documentation, treat the taxonomy as a checklist against your own templates. Do you name the same information elements? Do you map them to the same stakeholders? Where your terms and the standard's terms differ, decide whether to adopt its language or at least maintain a crosswalk so you can translate on request. Do not overreact and treat it as a legal requirement, because it is not one. Read it as a reference that makes your disclosures legible to people who use the same reference.

Questions professionals are asking

Does ISO/IEC 12792:2025 require companies to disclose anything?

No. It is a voluntary international standard that provides a taxonomy of transparency information elements. It has no legal force on its own. A disclosure duty would come from a contract, a certification scheme, or a law, not from the standard itself.

What does the standard actually contain?

Per the ISO catalogue, it specifies a taxonomy of information elements to help AI stakeholders identify and address transparency needs, and it describes the meaning of those elements and their relevance to different stakeholders. It is a classification and vocabulary, not a set of pass or fail rules.

Is this the same as ISO/IEC 42001?

No. ISO/IEC 42001 is an AI management-system standard. ISO/IEC 12792 is a narrower transparency taxonomy. They are separate documents. The taxonomy can support transparency work inside a management system, but it is not a management-system standard itself.

Why should a US professional care about a voluntary standard?

Because voluntary AI standards often get incorporated into contracts, certifications, and the way regulators describe reasonable practice. Aligning your documentation vocabulary with a common taxonomy early is cheaper than retrofitting it later when a customer or an obligation asks you to show organized transparency practices.

RELATED BRIEFINGS

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any standard, statute, or requirement applies to your situation with qualified professionals in the relevant jurisdiction.