---
id: sgkb.business-industry.securing-agentic-ai-addendum-2026
kind: article
title: "Securing Agentic AI Addendum 2026"
url: https://ik.sg/en/knowledge/article/sgkb.business-industry.securing-agentic-ai-addendum-2026
language: en
evidence_status: verified
last_verified: 2026-10-23
collection_as_of: 2026-11-22
---

# Securing Agentic AI Addendum 2026

Singapore's Cyber Security Agency published a 2026 addendum to its Guidelines and Companion Guide on Securing AI Systems, giving system owners agentic-AI risk-assessment methods and practical lifecycle controls for systems that can plan and act with access to tools and data.

# Securing Agentic AI Addendum 2026

Singapore's **Cyber Security Agency (CSA) Securing Agentic AI Addendum** was published on **17 June 2026** for public consultation as an addition to CSA's 2024 Guidelines and Companion Guide on Securing AI Systems. The CSA publication page, last updated **23 October 2026**, continues to describe the addendum as supporting system owners and being read alongside those foundational documents; the page retains the original consultation window and release date ([CSA, 17 June 2026; page last updated 23 October 2026](https://www.csa.gov.sg/resources/publications/addendum-on-securing-ai-systems/)).

## What makes agentic AI a distinct security issue

CSA describes agentic AI as able to understand context, formulate plans and take independent actions to achieve specified objectives. This creates new risks and potentially greater impact because such systems may access tools and data. Agentic status depends on system capabilities, not simply on whether a system uses AI. The addendum supplements foundational lifecycle security guidance; it does not establish a licensing regime ([CSA, 17 June 2026; accessed 10 October 2026](https://www.csa.gov.sg/resources/publications/addendum-on-securing-ai-systems/)).

The distinction is about system capability and autonomy, not a product label or industry sector. A system owner should first establish what the system can do, which tools and data it can reach, and how an agentic workflow can change outcomes; the addendum’s risk framing is consequently relevant to systems with different levels of autonomy. It should not be applied as proof that an ordinary rules-based application is automatically an agentic system ([CSA, Securing Agentic AI Addendum, 17 June 2026](https://www.csa.gov.sg/resources/publications/addendum-on-securing-ai-systems/)).

## Risk assessment and lifecycle controls

The addendum explains how system owners can identify and assess risk according to an agentic system's capabilities, including mapping agentic workflows to locate points where threat actors could potentially exploit vulnerabilities. It also provides practical controls to mitigate relevant risks across the development lifecycle, with examples across different scenarios and levels of system autonomy. CSA developed it with industry, government and international partners ([CSA, 17 June 2026; accessed 10 October 2026](https://www.csa.gov.sg/resources/publications/addendum-on-securing-ai-systems/)).

This makes the document a companion for security design and review rather than a one-time pass/fail test. The workflow-mapping approach helps a system owner relate threats to the system’s actual actions, tools and data access, while the lifecycle controls provide considerations for development and deployment stages. The page does not promise that following the examples eliminates risk or that CSA has assessed a particular company’s implementation ([CSA, Securing Agentic AI Addendum, 17 June 2026](https://www.csa.gov.sg/resources/publications/addendum-on-securing-ai-systems/)).

## Examples and status of the guidance

CSA gives app-development and coding assistants, automated client-onboarding systems and automated fraud-detection systems as example use cases, not an exhaustive scope or certification checklist. The page says the public consultation ran from **22 October to 31 December 2025**, and that the addendum was released on 17 June 2026 to support system owners. This is practical guidance rather than a universal certification or statutory approval; the examples do not represent CSA assessment of any particular organisation's implementation ([CSA, 17 June 2026; accessed 10 October 2026](https://www.csa.gov.sg/resources/publications/addendum-on-securing-ai-systems/)).

## Scope and quantitative evidence boundary

The CSA publication is a guidance document, not a programme that enrolls organisations or reports a count of compliant systems. Its official release page provides no number of adopters, assessed deployments, vulnerabilities prevented, consultation respondents or success rate. The page does not define an eligible-population denominator from which an adoption or compliance percentage could be calculated. The 2025 consultation dates and 17 June 2026 publication date describe the document's process and release, not participation volume, commencement of a legal requirement, or a measured security outcome ([CSA, 17 June 2026; accessed 10 October 2026](https://www.csa.gov.sg/resources/publications/addendum-on-securing-ai-systems/)).

## Sources

- [Cyber Security Agency — Securing Agentic AI: An Addendum to the Guidelines and Companion Guide on Securing AI Systems](https://www.csa.gov.sg/resources/publications/addendum-on-securing-ai-systems/) — Accessed 2026-10-23

## Explore connections

- Articles: [CSA CISOaaS cybersecurity services 2026](https://ik.sg/en/knowledge/article/sgkb.business-industry.csa-cisoaas-cybersecurity-services-2026)
- Articles: [CSA Cybersecurity Codes of Practice for CII and Cloud Services](https://ik.sg/en/knowledge/article/sgkb.business-industry.csa-ccop-cii-cloud-2026)
- Articles: [Cyber Security Agency of Singapore](https://ik.sg/en/knowledge/article/sgkb.government-politics.cyber-security-agency)
- Articles: [Google–Singapore AI Agents Sandbox](https://ik.sg/en/knowledge/article/sgkb.business-industry.google-ai-agents-sandbox)
- Articles: [Singapore's National AI Strategy and Governance](https://ik.sg/en/knowledge/article/sgkb.business-industry.national-ai-strategy-and-governance)
- Datasets: [Securing Agentic AI Addendum — Publication and Scope Anchors](https://ik.sg/en/knowledge/dataset/sgkb.data.securing-agentic-ai-addendum-anchors)
- Topics: [Business and industry](https://ik.sg/en/knowledge/topic/sgtopic.business)
- Topics: [Securing Agentic AI Addendum 2026](https://ik.sg/en/knowledge/topic/sgtopic.business.securing-agentic-ai-addendum-2026)
- People & organisations: [Cyber Security Agency of Singapore](https://ik.sg/en/knowledge/entity/sg.agency.csa)
- People & organisations: [Securing Agentic AI Addendum 2026](https://ik.sg/en/knowledge/entity/sg.framework.securing-agentic-ai-addendum-2026)

Dates describe this record’s own period and applicability. A verification date does not mean a rule is currently in force. An expanding collection. Published counts show available knowledge, not complete coverage of Singapore.
