TR 147:2026

Cybersecurity practices for large language model applications

OVERVIEW

This Technical Reference applies to organisations involved in the development, deployment, integration or maintenance of LLM-based systems, whether in-house or outsourced. It addresses the distinct security challenges of LLMs arising from their dependencies on datasets, model artefacts, open-source toolchains, infrastructure components and integration environments.

It excludes non-security aspects of AI trustworthiness, such as ethics, transparency and explainability, except where these considerations directly intersect with security.

It provides recommendations that can be adapted to different system scales, operational contexts and risk tolerances, forming a flexible component of organisation-specific AI risk-management and governance frameworks.

COMMENTS

-

PRODUCT DETAILS

Status Current
Edition 2026
No. of Pages 85
ICS Classification 35.020 Information technology (IT) in general
35.030 IT Security
Committee Information Technology Standards Committee
Available for Purchase Global
Adoption -