SA TS ISO IEC 8200-2024 PDF

Full title and description

St SA TS ISO IEC 8200-2024 — ISO/IEC TS 8200:2024, Information technology — Artificial intelligence — Controllability of automated artificial intelligence systems. This technical specification defines a framework of principles, characteristics and approaches to realise and enhance controllability in automated AI systems, with guidance applicable across the AI system lifecycle.

Abstract

This technical specification specifies a basic framework with principles, characteristics and approaches for the realisation and enhancement of controllability for automated AI systems. It addresses state observability and state transition, control transfer processes and associated costs, reactions to uncertainty during control transfer, and verification and validation approaches. The guidance is intended for all organisations developing or using AI systems throughout their lifecycle.

General information

• Status: Published (Technical Specification).
• Publication date: 10 April 2024.
• Publisher: ISO/IEC (published under ISO/IEC JTC 1/SC 42).
• ICS / categories: 35.020 (Information technology).
• Edition / version: Edition 1 (2024), TS.
• Number of pages: 34 pages.

Scope

The scope covers a foundational controllability framework for automated AI systems, including: mechanisms for observing internal and external system state; models and rules for state transitions; structured processes for transferring control between automated components and human operators (including cost and timeliness considerations); approaches to handle uncertainty and partial observability during control transfer; and verification, validation and testing methods to demonstrate and maintain controllability across the AI lifecycle. It is intended to be applicable across sectors and organisation types.

Key topics and requirements

• Definitions and principles of controllability for automated AI systems (terminology aligned with ISO/IEC AI vocabulary).
• State observability and instrumentation: what to monitor and how to represent system state for control purposes.
• State transition modelling: expected vs. unexpected transitions and modelling of control authority boundaries.
• Control transfer processes: handover protocols, timing, escalation, human–machine interface considerations and cost assessment.
• Managing uncertainty during transfer: detection, mitigation and fallback strategies when observability is incomplete.
• Verification, validation and testing approaches to assess controllability (including metrics, test scenarios and continuous monitoring).
• Documentation, governance and role/responsibility allocation to support operational controllability and audits.

Typical use and users

Intended users include AI system architects and engineers, safety and assurance teams, product managers, system integrators, risk and compliance officers, procurement and legal teams specifying controllability requirements, and regulators or evaluators assessing AI deployments. Typical uses include designing human-in-the-loop/human-on-the-loop interfaces, creating handover procedures for autonomous systems, defining verification and validation plans, and embedding controllability requirements into procurement and governance frameworks.

Related standards

Belongs to the ISO/IEC JTC 1/SC 42 AI family. Closely related documents include ISO/IEC 22989 (AI concepts and terminology), ISO/IEC TR 24028 (overview of trustworthiness in AI), ISO/IEC 23894 (guidance on AI risk management), and ISO/IEC 42001 (AI management system). These provide foundational terminology, trustworthiness and risk-management guidance, and management-system requirements that complement controllability guidance in TS 8200.

Keywords

controllability, control transfer, state observability, state transition, human-in-the-loop, verification and validation, AI assurance, automation handover, SC 42, AI lifecycle, reliability, safety.

FAQ

Q: What is this standard?

A: ISO/IEC TS 8200:2024 is a Technical Specification that defines a framework and practical guidance for ensuring and assessing the controllability of automated AI systems — i.e., the ability to observe, understand and transfer control safely and predictably between automated functions and human operators.

Q: What does it cover?

A: It covers principles, characteristics and approaches for state observability, state-transition modelling, control-transfer processes (including cost and timing), handling uncertainty during transfer, and verification/validation activities to demonstrate controllability across the AI lifecycle.

Q: Who typically uses it?

A: Engineers and architects designing AI systems, safety and assurance teams, risk and compliance staff, product and programme managers, procurers and regulators — any stakeholders responsible for design, deployment, operation, oversight or evaluation of automated AI systems.

Q: Is it current or superseded?

A: As published on 10 April 2024, ISO/IEC TS 8200:2024 is the current Technical Specification; it is not listed as superseded. Users should check national or ISO catalogues for any subsequent amendments or replacements (this document is a TS and may be complemented by later standards or normative guidance).

Q: Is it part of a series?

A: Yes. It is part of the ISO/IEC JTC 1/SC 42 portfolio of AI standards and complements other AI standards addressing terminology, trustworthiness, risk management and AI management systems.

Q: What are the key keywords?

A: controllability, control transfer, observability, state transition, verification, validation, human oversight, AI assurance, automation safety.