PDPC Issues Advisory Guidelines on Personal Data in Generative AI
PDPC issued Advisory Guidelines on Use of 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 Generative AIArtificial Intelligence on 20 July 2026. They cover scraping behind digital barriers, AIArtificial Intelligence-specific notices for training on user data, and deployer responsibility. They are advisory, and the PDPA prevails.
What happened
Singapore’s Personal Data Protection Commission issued Advisory Guidelines on Use of Personal Data in Generative AI on 20 July 2026. Paragraph 1.5 says they are advisory and not legally binding, and that the PDPA prevails over them. They do not limit the Commission’s enforcement of the PDPA.
What it changes
The main positions by stage:
- Scraping: where online data is accessible without restriction, organisations can rely on the Publicly Available Exception, subject to the reasonable person test (paragraph 3.3).
- Digital barriers: for data behind paywalls, registration, authentication, geo-blocks or bot blockers, organisations must assess that the data is publicly available before relying on the exception (paragraph 3.4).
- User data: general notifications are an insufficient means of obtaining consent for large-scale model training or fine-tuning. Organisations must provide AIArtificial Intelligence-specific notifications (paragraph 4.4).
- Deployers: system deployers bear primary responsibility for ensuring the systems they choose can meet their PDPA obligations (paragraph 9.1).
- Requests: paragraph 10.5 lists best practices for access and correction, including data provenance records for training data.
Who is affected
Organisations that develop generative AIArtificial Intelligence models or systems in Singapore, or that deploy them. The Guidelines name model providers, system providers and system deployers.
What to do now
- List where your training data comes from, and which sources sit behind a digital barrier.
- Check whether you use customer data for training, and replace general notices with AIArtificial Intelligence-specific ones.
- Ask your AIArtificial Intelligence suppliers for the safeguard information you need to assess a system before you buy it.
- Keep provenance records for training data.
AI Governance Officer as-a-Service
We map your generative AIArtificial Intelligence training data against the digital barrier test, rewrite your notices as AIArtificial Intelligence-specific notifications, and set up the supplier checks you run as a system deployer.
