Singapore Studies Reporting of AI Incidents and Near-Misses, No Mandate Yet

Singapore's digital ministry says IMDA and the Cyber Security Agency are studying how to report serious AI incidents and near-misses, but it announced no mandate, no timeline and no new rule.

By Utopia Data & AI Team5 min read

Grey Parliament House building with terracotta roofs and red banners behind palm trees and a lawn, with office towers rising behind it
Licensed photo. Parliament House in Singapore, photographed on 13 August 2023. The image is contextual and shows nothing from the 4 November sitting. Photo: Johan Jönsson/Wikimedia Commons, CC BY-SA 4.0.[7][1]

In brief

  • On 4 November 2026 Singapore's digital ministry answered MPs' questions on AI risk in Parliament. It said the Infocomm Media Development Authority (IMDA) and the Cyber Security Agency (CSA) are studying how to strengthen reporting of serious AI-related incidents.[1]
  • The answer does not say reporting will become mandatory. It says any extra arrangement should be targeted and proportionate, and it gives no date for a decision.[1]
  • Three other answers the same day covered self-replicating AI agents, a possible AI assurance industry and entry-level technology hiring. Each cites existing programmes rather than new rules.[2][3][4]

What the ministry said

Singapore's Ministry of Digital Development and Information (MDDI) published its answers to four parliamentary questions on AI on 4 November 2026. The questions were put by MPs Melvin Yong Yik Chye, Sharael Taha and Joan Pereira. The ministry's pages say the oral questions were converted to written answers, so no debate is recorded in the documents we read.[1][2][3][4]

The most closely watched answer concerns incident reporting. Two MPs asked whether Singapore will require reports of serious incidents and near-misses involving autonomous AI systems. The ministry did not say yes or no. It said IMDA and CSA are studying how to strengthen reporting, including whether new reporting or information-sharing arrangements are needed.[1]

Singapore's financial district towers behind a white clock tower and a colonial-era building with red banners, under a clear blue sky
Licensed photo. Singapore's Central Business District seen from the roof of the National Gallery, photographed on 8 June 2018. The image is contextual and shows no AI system. Photo: Basile Morin/Wikimedia Commons, CC BY-SA 4.0.[8][1]

The reporting gap the ministry describes

The ministry says Singapore already requires reports of serious incidents involving personal data, cybersecurity and disruption to critical or regulated services. Those rules apply whether or not AI played a part. It then names a gap. Autonomous AI systems may take unintended actions, and such incidents may fall below existing thresholds, particularly when they cause no reportable harm.[1]

In the ministry's view, information about near-misses could still show emerging risks and help build safeguards before harm occurs. It said IMDA and CSA will draw on international approaches and build on existing mechanisms, such as CSA's exchange of cybersecurity intelligence with major technology providers and critical information infrastructure operators, where feasible.[1]

What is not decided

The answer sets no date, names no legal instrument and gives no definition of a near-miss. It says any additional arrangement should be targeted and proportionate and should consider legitimate security and commercial sensitivities. Singapore will also work with interested partners on a common understanding of AI incidents and near-misses, and on ways to share information securely.[1]

Readers should treat this as a study, not a rule. A study can end in a mandatory regime, a voluntary scheme or no change. The ministry's wording leaves all three open, and we have no source that says which is likelier.[1]

Perdana Putra in Putrajaya, with a green central dome, Malaysian flags on its facade and an empty paved approach road in rain
Licensed photo. The Perdana Putra building in Putrajaya, photographed on 4 August 2024. NACSA's page lists its office in this building. The image is contextual and shows no AI incident. Photo: Stress 043/Wikimedia Commons, CC BY-SA 4.0.[9][6]

Three other answers on 4 November

On AI agents that copy themselves across devices, the ministry said self-replicating malware is not new. It said recent research has shown AI agents can replicate under pre-set conditions, but that does not mean they can easily spread in real-world settings. It pointed to CSA's Cybersecurity Labelling Scheme, the SG Cyber Safe programme, the Cyber Essentials mark and a CISO-as-a-Service programme for small businesses.[2]

On AI assurance, the ministry said demand for AI testing services will grow and the industry's shape remains unknown. It cited the AI Verify Foundation's tester accreditation programme and Singapore's proposal for ISO/IEC 42119-8 on testing generative AI. On hiring, it said IMDA tracks tech workforce trends through two surveys, is co-leading a workgroup on the future of tech work, and expects employers to adapt how they train junior staff.[3][4]

Why Malaysia should care

Singapore's answers do not mention Malaysia. NACSA says the Cyber Security Act 2024 covers cyber threats and incidents involving National Critical Information Infrastructure. We have not read the Act to see how it treats an AI near-miss, so Malaysian teams should not assume a reporting duty either way. Companies that use AI agents can still write down near-misses now.

Malaysian firms that run AI agents with access to data or systems

Singapore is studying whether near-misses should be reported. It has not said they must be.[1]

Practical move: Start a simple internal log of unintended agent actions.

Malaysian IT and security teams serving Singapore clients

A Singapore reporting scheme, if one comes, could reach client contracts. No scheme exists yet.[1]

Practical move: Ask clients whether any AI incident clause is planned.

Malaysian graduates and employers in tech

Singapore expects employers to change how they train junior staff as AI changes entry-level work.[4]

Practical move: Review which junior tasks AI now covers and how juniors will learn.

What Malaysians can do now

  1. Keep a written log of any unintended or near-miss action by an AI agent, even when no harm results.
  2. Check whether Malaysia's Cyber Security Act 2024 applies to your sector before assuming an AI incident duty exists.
  3. Read CSA's agentic AI addendum and map which of your agent workflows can reach sensitive data or systems.

What we still do not know

What we still do not know

  • Whether Singapore will make AI incident and near-miss reporting mandatory, and by when.
  • How a near-miss would be defined, and who would have to report it.
  • Whether Malaysian regulators are studying a similar step, since NACSA's page says nothing on AI near-misses.

Sources

  1. MDDI's Response to PQ on Reporting of Serious and Near-miss Incidents Involving Autonomous AI SystemsMinistry of Digital Development and Information, Singapore, 4 November 2026
  2. MDDI's Response to PQ on Assessment of Risk of AI Systems Replicating Across Internet-connected Devices in SingaporeMinistry of Digital Development and Information, Singapore, 4 November 2026
  3. MDDI's Response to PQ on Establishing Singapore as Global Leader in Agentic AI Security and AssuranceMinistry of Digital Development and Information, Singapore, 4 November 2026
  4. MDDI's Response to PQ on Changes in Entry-level Hiring for Software Engineering and Other Technology Role Arising from AI AdoptionMinistry of Digital Development and Information, Singapore, 4 November 2026
  5. Securing Agentic AI – An Addendum to the Guidelines and Companion Guide on Securing AI SystemsCyber Security Agency of Singapore, 17 June 2026
  6. Cyber Security Act 2024 (Act 854)National Cyber Security Agency (NACSA), Malaysia
  7. File:Parliament House, Singapore, August 2023.jpgWikimedia Commons, 13 August 2023
  8. File:Skyline of the Central Business District with the Old Parliament House in Singapore.jpgWikimedia Commons, 8 June 2018
  9. File:Front building Perdana Putra Putrajaya 2024.jpgWikimedia Commons, 4 August 2024

Facts come from four MDDI answers dated 4 November 2026, CSA's addendum page and NACSA's Cyber Security Act page, all read directly on 15 November. We did not read the CSA addendum itself or the text of Malaysia's Act, and we did not contact MDDI, IMDA or CSA. The three photographs are licensed contextual images and none shows the events described. No generated image is used.

Last fact check: 15 November 2026. Factual correction? devteam.utopia@gmail.com.

More AI News

Back to AI NewsAll 201 stories