AI assistants side by side for Singapore property agents: ChatGPT, Gemini, Claude and Copilot on client data

Key Takeaways

If you are a Singapore property agent, what happens to the client details you paste into an AI assistant depends on the plan you use more than on the brand: personal plans can use your chats to improve models unless you change a setting, while the business plans of all four providers say they do not train on your data by default.

• ChatGPT Free, Plus and Pro may use your conversations to train models; you can turn this off under Settings, Data controls, “Improve the model for everyone” (OpenAI, checked 20 October 2026).
• Google says human reviewers read some Gemini Apps chats and asks you not to enter confidential information; Gemini inside a Google Workspace account is not used for model training without permission (Google, checked 20 October 2026).
• Claude’s personal plans use chats to improve models when Model Improvement is switched on; its commercial products are not used for training by default (Anthropic, checked 20 October 2026).
• Microsoft says prompts and responses in Microsoft Copilot for organisations are not used to train foundation models (Microsoft, checked 20 October 2026).
• On every plan, a thumbs up or thumbs down can send the whole conversation to the provider.

In Detail

Most agents start with whichever AI assistant is already on their phone. That is usually a personal plan, and personal plans are where the training settings matter. Singapore’s Personal Data Protection Commission (PDPC) has published Advisory Guidelines on Use of Personal Data in Generative AI, so it is worth knowing what each tool does with what you type before a client’s name, NRIC number or financial details go in.

How we compared We looked at ChatGPT, Gemini, Claude and Copilot on four questions: whether a personal plan can use your chats for training, how to switch that off, what the business plan says, and whether the provider mentions human review. Each cell comes from the provider’s own help or privacy page, checked on 20 October 2026. An empty cell means we found no first-party statement on that point, not that the feature is missing.

Personal plans are where the risk sits

OpenAI says conversations on ChatGPT Free, Plus and Pro may be used to train its models, and that you can opt out under Data controls or in its privacy portal (OpenAI Help Center). Google’s Gemini Apps privacy page says trained reviewers, including from its service providers, read some chats, and asks users not to enter anything confidential they would not want a reviewer to see; chats that were reviewed are kept for up to three years even if you delete your activity (Gemini Apps Privacy Hub). Anthropic says chats on Claude’s personal plans are used to improve its models when Model Improvement is enabled, and that Incognito chats are not used (Anthropic Privacy Center, consumer).

Business plans say no training by default

The business versions read differently. OpenAI says it does not use inputs or outputs from ChatGPT Business, Enterprise, Edu or its API to improve models by default. Google says Workspace does not use customer data, including prompts, for model training without the customer’s prior permission (Generative AI in Google Workspace Privacy Hub). Anthropic says its commercial products are not used for training by default (Anthropic Privacy Center, commercial). Microsoft says prompts, responses and data accessed through Microsoft Graph in Copilot for organisations are not used to train foundation models (Microsoft Learn: enterprise data protection).

The feedback button is the exception

All three of OpenAI, Google and Anthropic describe feedback separately. OpenAI says that on personal plans a thumbs up or down can send the whole conversation for training even after you opt out. Anthropic says feedback stores the entire related conversation for up to five years, on personal and commercial plans. If a chat contains client details, do not rate it.

What the plan does not change

A business plan changes what the provider does with your data. It does not change what you are allowed to collect or disclose about your client. The safest habit is the same on every plan: replace names, NRIC numbers, addresses and figures with placeholders before you paste, and add them back yourself. The Cyber Security Agency of Singapore has also published Guidelines and Companion Guide on Securing AI Systems for organisations that build or run AI tools.

Putting This Into Practice

This belongs to the DELEGATE step of Guardrails.

1. Check which plan you are on. Free, Plus, Pro or Max are personal plans; Business, Enterprise, Workspace and “for Work” plans are business plans.
2. Open the data settings today. ChatGPT: Settings, Data controls. Gemini: Gemini Apps Activity. Claude: Privacy Settings. Decide whether training stays on.
3. Use a temporary or incognito chat for anything that touches a client.
4. Strip identifiers first. Placeholders for names, NRIC numbers, unit numbers and money.
5. Do not rate client chats with thumbs up or down.
6. Read every draft and approve it before it goes to a client.

Questions to check on each provider’s own site before you choose: Is my plan personal or business? Is training on by default? Where do I switch it off? Does feedback send the whole chat? How long are chats kept?

Key Questions Answered

Does ChatGPT train on what I paste in?

On ChatGPT Free, Plus and Pro it may, unless you turn off “Improve the model for everyone” in Settings, Data controls. OpenAI says ChatGPT Business, Enterprise, Edu and API data is not used to improve models by default (checked 20 October 2026).

Do people at Google read my Gemini chats?

Google says trained reviewers read some Gemini Apps chats and asks you not to enter confidential information. Gemini in a Google Workspace account is not used for model training without permission (checked 20 October 2026).

Is Claude safer for client data than ChatGPT?

Not by brand alone. Both have personal plans that can use chats when a setting is on, and business plans that are not used for training by default. Compare the plan you would actually use.

Does Microsoft Copilot use my prompts to train its models?

Microsoft says prompts and responses in Microsoft Copilot for organisations are not used to train foundation models. We did not check the personal Copilot app for this comparison.

Is it enough to switch training off?

No. Switching training off changes what the provider does with your chat; it does not make it right to paste a client’s personal data. Replace identifiers with placeholders first, and check the PDPC’s advisory guidelines on personal data in generative AI.

Sources

1. OpenAI Help Center, How your data is used to improve model performance, help.openai.com (checked 20 October 2026)
2. Google, Gemini Apps Privacy Hub, support.google.com/gemini (checked 20 October 2026)
3. Google, Generative AI in Google Workspace Privacy Hub, support.google.com/a (checked 20 October 2026)
4. Anthropic Privacy Center, Is my data used for model training? (consumer), privacy.anthropic.com (checked 20 October 2026)
5. Anthropic Privacy Center, Is my data used for model training? (commercial), privacy.claude.com (checked 20 October 2026)
6. Microsoft Learn, Enterprise data protection in Microsoft Copilot and Microsoft Copilot Chat, learn.microsoft.com (checked 20 October 2026)
7. Personal Data Protection Commission, Advisory Guidelines on Use of Personal Data in Generative AI, pdpc.gov.sg
8. Cyber Security Agency of Singapore, Guidelines and Companion Guide on Securing AI Systems, csa.gov.sg