How to use AI on your work phone without leaking client data

You can use AI on your work phone more safely, but only if you change the default privacy settings and keep client data out of prompts. The safest rule is simple: use on-device features first, turn off chat saving or training where possible, and never paste names, ID numbers, account details, contracts or confidential files into a consumer AI app.

If a task includes client records, legal drafts, HR issues, medical details or bank information, do not use a standard consumer AI app unless your employer has approved that exact tool and setup. For most business users, AI is best for low-risk drafting, summaries and rewrites using redacted text.

Decide by risk level before you pick an AI app

Do not start with the brand name. Start with the question: what data will leave your phone, and does the service keep or use that data?

Galaxy AI, Apple Intelligence, Gemini and ChatGPT do not all work the same way. Some features run on the phone, some rely on the cloud, and some keep activity unless you change a setting.

Use this simple risk test

• Low risk: rewriting your own generic email, summarising public notes, drafting an agenda with no client details.
• Medium risk: internal notes, non-public dates, project names, pricing discussions, staff information.
• High risk: client records, bank details, South African ID numbers, contracts, health data, legal advice, HR complaints, confidential attachments.

Low-risk tasks may be fine in a consumer AI app with strict privacy settings. High-risk tasks should stay out of consumer AI unless your company has an approved enterprise setup and policy.

Use on-device AI first on Samsung and iPhone

If privacy matters most, on-device features are your best starting point. They reduce the chance that your text, voice or files leave the phone for cloud processing.

That does not make every feature safe for client work. You still need to check each setting and avoid feeding in confidential material.

Samsung Galaxy AI: check each feature one by one

Samsung mixes on-device and cloud-based AI features. On a phone such as the Galaxy S25, which starts from R15,477 in our latest checks, you should review every Galaxy AI feature before using it for work.

1. Open Settings.
2. Tap Galaxy AI or search for the feature name.
3. Review tools such as writing help, summaries, call assistance and transcription.
4. Turn off online processing where the option is available and you do not need it.
5. Read the privacy notice inside the feature before using work content.

A premium phone does not automatically mean enterprise-grade privacy. Your protection depends more on the feature settings and your prompt habits than on the hardware.

Apple Intelligence: private by design, but still not a free pass

Apple positions Apple Intelligence around on-device processing and private handling. That makes it easier to justify for low-risk work tasks, but it still does not mean every request stays on the phone.

If you use an iPhone 17e, from R16,355 in our tracked pricing, or a higher-end iPhone such as the iPhone 18 Pro Max from R30,699, check what each feature does before you use business text.

1. Open Settings.
2. Check Apple Intelligence and Siri.
3. Turn off options that share recordings or interactions for service improvement, if available.
4. Review app permissions for Photos, Files, Contacts and Microphone.
5. Use AI writing tools only with redacted text unless your company says otherwise.

The safest view is this: Apple Intelligence and some Galaxy AI features are better suited to low-risk work than copying raw client material into a cloud chatbot. They are still not the place for unredacted contracts, IDs or medical records.

Tighten Gemini and ChatGPT before you use them for work

Gemini and ChatGPT are useful, but they are also where many people make their biggest privacy mistakes. They are fast, easy to reach, and tempting when you need to rewrite a message or summarise notes in a hurry.

That convenience is exactly why you should lock them down first.

Gemini on Android: treat it as cloud-connected unless clearly told otherwise

Gemini is deeply tied into Android and your Google account. That can be helpful, but it also means you should be careful about linked apps, saved activity and access to work data.

1. Open the Gemini app or Gemini settings in your Google account.
2. Look for Gemini Apps Activity or similar activity controls.
3. Pause or turn off activity saving if your company policy allows it.
4. Review connected apps and remove access you do not need.
5. Do not connect Gemini to work email, files or notes unless your employer has approved it.

If you use one Android phone for both personal and work life, separate your accounts properly. If you are comparing phones, our phone comparisons page is a better place to start than AI marketing claims.

ChatGPT on mobile: turn off chat saving or training first

ChatGPT is often the first AI app people install for work help. It is also the app most likely to receive copied client text, screenshots and uploaded files.

1. Open the ChatGPT app.
2. Tap your profile or menu.
3. Open Settings.
4. Go to Data Controls or the privacy section.
5. Turn off Chat History & Training or similar options if they are available on your account.
6. Review permissions and remove access to Photos, Microphone, Contacts and Files unless you need them.
7. Do not upload client documents, screenshots or voice notes.

Use ChatGPT for templates, checklists and rewrites of anonymised text. For money-related questions, our piece on whether you should trust ChatGPT with money explains why caution matters.

What you should never paste into a consumer AI app

The biggest mistake is not the app. It is the prompt.

If the text would be a problem in the wrong inbox, do not put it into a consumer AI tool on your phone.

• South African ID numbers or passport numbers
• Bank account or card details
• Client contracts and signed PDFs
• Medical records or insurance claims
• HR complaints, disciplinary notes or salary data
• Private WhatsApp screenshots from clients or staff
• Customer account numbers tied to names or addresses

Worked example: unsafe vs safer prompt

Unsafe example: “Summarise this complaint from Mr Dlamini at account 4582. He says Vodacom billed him R349 for a line he cancelled and his ID is 800101...”

Safer example: “Summarise this customer complaint in a professional tone. Remove names, account numbers, ID numbers and network references. Complaint: a customer says they were billed after cancelling a mobile line and wants a refund.”

The second version still gives you useful help, but strips out the details that could expose your client or your business.

Which AI option is safest for work on a phone?

If privacy is your top priority, on-device features are usually the safest option. Cloud-linked assistants can still be useful, but they need stricter settings and stricter habits.

For most South African professionals, Apple Intelligence and some Galaxy AI features are easier to justify for low-risk tasks than Gemini or ChatGPT with raw client data. If your employer has no AI policy yet, take the conservative route and keep anything client-specific out of the prompt.

Set a simple AI phone policy for yourself or your team

You do not need a long policy to cut most of the risk. You need a short set of rules people can follow on a busy day.

1. Use on-device AI first where possible.
2. Turn off chat saving or model training in consumer AI apps.
3. Never paste names, numbers, contracts or screenshots with personal data.
4. Use redacted prompts for drafting and summaries.
5. Keep work and personal accounts separate on Android and iPhone.
6. Ask your employer or IT team before linking AI to work email, files or contacts.

If you rely on mobile data, remember that cloud AI also uses data when it sends prompts and files. Our guides on cutting WhatsApp, Teams and Zoom data use and using less mobile data on your phone can help.

Frequently asked questions