PDPC says organizations should issue AI-specific notices rather than relying on broad privacy notices
[SINGAPORE] Organizations that use personal data to train generative artificial intelligence models must clearly notify affected users under the proposed new rules.
The Personal Data Protection Commission (PDPC), which consulted the public on the rule, said it was a necessary safeguard against the misuse of consumer data to train completely unanticipated services such as financial profiling.
In its proposed advisory guidelines, the PDPC said organizations should issue AI-specific notices rather than relying on broad privacy notices stating that personal data may be used for purposes such as “new product development.”
The proposed rules come as more companies, from banks and insurance companies to retailers and social media platforms, deploy their own generative AI models to personalize services and improve products.
“Individuals may have legitimate concerns about more sensitive personal data being published, reconstructed and/or disclosed to third parties from generative AI models,” the PDPC said in its consultation paper. The month-long public consultation ended on July 1st.
Personal data may include your name, email address, video and audio recordings, transaction history, and location data.
Visible AI-specific notifications can be in-product pop-ups or dedicated web pages. AI-specific notifications should include detailed information about the behavior of the generative AI model and the personal data it captures.
For example, if your company is developing text-to-speech functionality, you need to inform users that their voice recordings will be used to train an AI model. Companies should also explain how the recordings will help the model recognize speech patterns.
AI-specific notices must also provide detailed instructions for opting out or withdrawing consent for personal data to be used for AI training.
Related items
For example, a social media platform that develops AI models to generate text and images may update its privacy policy with the following notice: “When you interact with our AI-enabled features, the text, images, and audio you send may be used for product improvement purposes, such as training and running our AI models. This allows our AI-enabled features to generate more realistic images and audio patterns.”
Social media platforms may also provide details on their web pages, either through privacy updates within the platform that users can read or via an email containing a link to opt-out of AI training.
It is unclear whether users will need to explicitly consent to AI training, or whether simply allowing them to opt out will be sufficient.
It is also unclear whether mandatory AI-specific notifications apply when personal data is anonymized before being used for AI training.
It’s also unclear whether organizations such as banks, insurance companies and social media platforms can deny services to consumers if they refuse AI training.
The proposed guidelines are one of the key priorities of Dennis Wong, who took over as PDPC’s fifth commissioner in April 2026.
The authority was established in January 2013 to enforce Singapore’s data protection laws, investigate complaints and data breaches, and educate organizations and individuals about their rights and obligations.
During an interview with Straits Times On June 23, Mr Wong spoke about the need to ensure Singapore’s data protection framework is relevant in the era of generative AI, particularly the increasing use of AI-enabled devices that collect biometric data.
For example, AI-enabled smart glasses, smart watches, and palm-scan payment systems collect facial features, fingerprints, voice recordings, and palm vein patterns, exposing users to new risks.
In particular, AI smart glasses, including the Ray-Ban Meta and Oakley Meta devices, have raised eyebrows because it may not be possible to tell whether they have been secretly recorded. In the UK, a BBC investigation in January 2026 found that women were being harassed after they were secretly filmed using Ray-Ban meta glasses and the videos were posted online along with their personal information.
It has also fueled fears of high-tech cheating, as students can easily record exams in secret without knowing anyone is watching. In South Korea, two test takers were caught cheating on an English proficiency test using AI-equipped smart glasses in May 2026, and authorities banned the devices from the test venue.
The PDPC is therefore considering whether it needs to advise businesses on what constitutes meaningful consent for the use of such AI-enabled devices. “We are thinking about whether we need to be clearer about the circumstances in which the use of such devices is and is not acceptable,” Wong said. strait era
