PDPC Tells Organisations to Control Shadow AI in Software Development
Singapore's PDPC advises organisations to inventory the AIArtificial Intelligence tools staff use, set policies, review AIArtificial Intelligence-generated code and security-test it before deployment, to meet the PDPA's security duty in section 24.
What happened
On 30 September 2026 Singapore’s Personal Data Protection Commission published an advisory on data protection risks when staff use unauthorised AIArtificial Intelligence tools to build software or scripts. It ties the measures to section 24 of the PDPA, which requires reasonable security arrangements for personal dataAny information relating to an identified or identifiable natural person, the data subject. A person is identifiable if they can be identified directly or indirectly, for example by a name, an identification number, location data, an online identifier or factors specific to their identity..
What it changes
The PDPC describes an employee who used a personal AIArtificial Intelligence account to build an email workflow and put it into production without testing, exposing customers’ email addresses. It recommends:
- an inventory of all AIArtificial Intelligence tools in use, with purpose, data flows and users, alongside the ICTInformation and Communication Technology asset inventory;
- a risk assessment and written policies for AIArtificial Intelligence tools, and least-privilege access;
- review of AIArtificial Intelligence-generated code by competent staff, training, and logging of AIArtificial Intelligence inputs and outputs;
- code review, acceptance testing and security testing, such as vulnerabilityA weakness, susceptibility or flaw that a threat can exploit. NIS2 applies it to ICT products and services, the CRA to products with digital elements and DORA to any asset, system, process or control. Coordinated vulnerability disclosure and handling duties build on this definition. assessment, penetration testing or red teaming, before production and after major changes;
- no production personal dataAny information relating to an identified or identifiable natural person, the data subject. A person is identifiable if they can be identified directly or indirectly, for example by a name, an identification number, location data, an online identifier or factors specific to their identity. in non-production environments.
Who is affected
Organisations in Singapore whose staff use AIArtificial Intelligence tools to write code, scripts or automated workflows that touch personal dataAny information relating to an identified or identifiable natural person, the data subject. A person is identifiable if they can be identified directly or indirectly, for example by a name, an identification number, location data, an online identifier or factors specific to their identity..
What to do now
- Build an inventory of the AIArtificial Intelligence tools your staff use and block or approve them.
- Set a written policy and a review step for AIArtificial Intelligence-generated code before deployment.
- Test AIArtificial Intelligence-built systems for security before they go live.
AI Governance Officer as-a-Service
We would build your AIArtificial Intelligence tool inventory and usage policy, set the review step for AIArtificial Intelligence-generated code, and test AIArtificial Intelligence-built systems before they reach production.
