By the WahLiao desk · Last verified 1 October 2026
What you type into a Super Intelligence chatbot, the kind of AI behind ChatGPT, Claude and Gemini, is sent to and stored by the company that runs it. On most consumer plans it can also be used to train future models unless you switch that off. Each of the big three has a setting for this, and business plans generally do not train on your content by default. Singapore’s Personal Data Protection Act (PDPA) governs how organisations collect and use personal data, which matters most at work: if you paste a customer’s details into a chatbot, your employer is still responsible for them.
The simplest rule is also the most useful. Do not type anything into a chatbot that you would not put in an email to a stranger: NRIC numbers, passwords, one-time codes, bank details, medical records or a client’s file.
AI and your data: quick facts
| What is stored | Your chats and uploaded files, kept by the provider on its own servers |
| Training | Consumer plans may use your chats to train models unless you opt out |
| ChatGPT | Settings, Data Controls, turn off “Improve the model for everyone”; or use Temporary Chat |
| Gemini | Turn off “Keep Activity” |
| Claude | Settings, Privacy, turn off the model-improvement setting; or use an incognito chat |
| Business plans | Generally no training on your content by default |
| The law | The PDPA sets rules for organisations handling personal data |
The WahLiao Verdict
| Do today | Check the training setting on every tool you use. |
| Never paste | NRIC, passwords, codes, bank details, health records, client data. |
| At work | Use the tool your employer approves, on its business plan. |
| Remember | Opting out stops future training; it does not undo the past. |
| Deleting | Providers may keep deleted chats for a period; check the privacy policy. |
What happens to what you type
When you send a message, it travels to the provider’s servers, usually overseas, where the model reads it and writes a reply. The conversation is normally saved to your history so you can return to it. Depending on the provider and your settings, conversations may also be reviewed for safety and used to improve future models. Once something has been used in training, switching the setting off later protects your future chats, not the ones already used.
How to switch off training
| Tool | Where | What it does |
|---|---|---|
| ChatGPT | Settings, Data Controls, “Improve the model for everyone” | Stops chats being used for training; your history stays, and the setting applies on all your devices |
| ChatGPT Temporary Chat | Start a Temporary Chat | Not saved to your history and not used for training |
| Gemini | Activity settings, “Keep Activity” | Turning it off stops your Gemini activity being kept for improving Google’s models |
| Claude | Settings, Privacy, the model-improvement setting | Anthropic lists training as opt-out on its Free, Pro and Max plans, and off by default on Team and Enterprise |
| Copilot | Microsoft’s privacy settings | Controls have changed more than once; check the current options in your account |
Settings move around as apps are updated, so if a label has changed, search the provider’s help pages for “training” or “data controls”.
What the PDPA covers
The PDPA sets rules for organisations that collect, use and disclose personal data in Singapore: broadly, they must tell you why they want it, get your consent unless an exception applies, keep it safe, and let you see and correct what they hold. It does not govern what you do as a private individual in your personal or domestic life, so your personal chats with a chatbot are a matter of the provider’s own terms and privacy policy.
In March 2024, the Personal Data Protection Commission issued advisory guidelines on how the PDPA applies when organisations use personal data in AI systems that make recommendations or decisions, such as when consent is needed, what customers should be told, and when exceptions for business improvement and research apply. The guidelines are not legally binding and focus on those recommendation and decision systems rather than on generative chatbots, but they show how the regulator thinks about AI and personal data.
At work: the risky shortcut
The most common data mistake is a well-meant one: pasting a customer list, a patient note or a contract into a personal chatbot account to save time. Under the PDPA, the organisation remains responsible for protecting that personal data. Use only the tools your employer has approved, preferably on a business plan that does not train on your content by default, and follow your company’s AI policy. If there is no policy, ask before you paste.
Photos, voices and your children
Uploading photos or voice recordings of other people, especially children, deserves extra thought. Images and audio can be kept, reused or leaked, and clips of voices and faces are the raw material of deepfake scams. Ask before you upload someone else’s picture, and keep children’s images off public accounts where you can.
If something goes wrong
If you think an organisation in Singapore has mishandled your personal data, raise it with the organisation first, then with the Personal Data Protection Commission. You can also ask an organisation what personal data it holds about you. For deleting your chat history with a chatbot, use the delete or export options in its settings, and read its privacy policy to see how long deleted data is kept.
AI and your privacy: FAQ
Does ChatGPT use my chats for training?
On consumer plans it can, unless you turn off “Improve the model for everyone” in Data Controls. Temporary Chats are not used for training.
Is it safe to type my NRIC number into a chatbot?
No. Keep NRIC numbers, passwords, one-time codes and bank details out of chatbots altogether.
Does the PDPA apply to AI?
It applies to organisations using personal data, including in AI systems. The PDPC’s 2024 advisory guidelines explain how it applies to AI recommendation and decision systems.
Do business AI plans train on my data?
Generally not by default. Anthropic, for example, lists no training by default on Claude’s Team and Enterprise plans.
If I delete a chat, is it gone?
It disappears from your history, but providers may keep deleted data for a period for safety or legal reasons. Check each provider’s privacy policy.
Read next
This page belongs to Super Intelligence. Next, read Singapore’s Super Intelligence plan.
Sources checked 1 October 2026: OpenAI Help Center, data controls and Temporary Chat; Anthropic, Claude plans and model-training settings; SCELC, how to disable AI model training (Gemini); PDPC, Advisory Guidelines on the use of personal data in AI recommendation and decision systems; Data Protection Report, on the guidelines’ scope. General information, not legal advice. Last updated 1 October 2026.
