"Is it safe?" is the first question I get from clinics, property managers and professional offices thinking about AI. It's the right question, and the honest answer is: it depends on which tool, which plan and what data. Here's how I'd think it through.
This is practical guidance, not legal advice. If you handle sensitive information, talk to your privacy officer or lawyer before you change how it's processed.
Consumer plans and business plans are not the same
The biggest difference isn't between ChatGPT, Claude or Gemini. It's between the free or personal version of a tool and the business version.
- Personal and free plans are built for individuals. Depending on the provider and your settings, what you type may be kept and may be used to improve future models. You often have to find and change a setting to opt out.
- Business plans and developer APIs are generally set up so your data isn't used for training by default, with contractual terms about retention, security and access. They usually cost more, and that's a large part of what you're paying for.
The practical rule: if staff are pasting client information into a personal account, that's the first thing to fix, whatever else you decide. Read the data terms for the specific plan you pay for, not the provider's marketing page.
What Alberta's privacy law expects
Most private businesses in Alberta fall under the Personal Information Protection Act (PIPA). In plain terms, it expects you to:
- Collect and use personal information only for reasonable purposes that people know about.
- Protect it with reasonable security arrangements.
- Tell people in your privacy policy when a service provider outside Canada handles their personal information. Most major AI services process data in the United States, so this often applies.
- Notify the Information and Privacy Commissioner when a breach creates a real risk of significant harm.
Using an AI service is, in effect, using a service provider. The same thinking you'd apply to cloud accounting software or an outsourced bookkeeper applies here.
If you're a health custodian under Alberta's Health Information Act, stricter rules apply to health information, and you should treat any AI use involving patient information as a formal privacy decision, not a staff productivity tip.
Sort your data into three groups
Before deciding on tools, I sort a client's information into three groups.
Fine to use with a business-grade AI service. Internal procedures, product information, public pricing, marketing drafts, anonymized or summarized data, and most routine business documents without personal details.
Allowed with care. Customer names and contact details, invoices, contracts, booking records. These can be processed by a properly configured business service, with the right privacy notice, limited access and a clear reason. Send only the fields the task needs.
Stays inside your systems. Health information, government ID numbers, banking and card details, anything covered by a confidentiality agreement that forbids it, and anything you'd struggle to explain to a client if it leaked. For this data, the AI comes to your systems, not the other way round, or it isn't used at all.
Design choices that reduce the risk
A lot of risk can be designed out:
- Send less. If the task is to categorize an invoice, the model needs the line items, not the customer's home address.
- Keep payment data out entirely. The phone agents I build never take card numbers. They send a secure link instead.
- Use your own accounts. When a client's data can't leave their environment, I work inside their systems and their AI platform, under their access controls.
- Log and review. Keep a record of what was processed and who saw the results.
- Guard against tricks. Documents and messages can contain hidden instructions aimed at the AI. Systems that read customer-supplied content need protections against that, like the injection guard on the ID reader I built for a rental counter.
Questions to ask any AI vendor
- Is my data used to train your models? Where is that written in the contract?
- Where is the data processed and stored, and for how long?
- Who on your side can access it, and when?
- Can I delete it, and how do I confirm that it's gone?
- What happens in a breach, and how quickly will you tell me?
If a vendor can't answer clearly, that's your answer.
The short version
AI can be used safely with client data, and plenty of businesses do. But it takes the same care as any other place your data goes: business-grade accounts, a clear privacy notice, sending only what's needed, and a firm line around the information that should never leave your hands. Decide where that line is before anyone starts pasting.