Is Perplexity Safe? Privacy for Tax Research

It depends: Perplexity can support tax research safely when you keep client facts out and use the privacy settings your practice approves.

Is Perplexity Safe for Tax Research?

Use Perplexity for questions about published tax guidance, with the client’s facts kept in your practice files. Your plan changes how Perplexity can use what you send.

Free, Pro and Max accounts have AI Data Retention enabled by default. That setting allows data collection for model training. Paying for a personal subscription does not give you the Enterprise data terms.

Enterprise Pro and Enterprise Max exclude customer data from AI training. They are the better fit when your practice needs organisation settings and contractual protection. Keep research questions free of client details on these plans too.

A Division 7A research question rarely needs the borrower’s identity or actual loan ledger. Describe the rule you want to understand, then apply the answer to the client’s records inside your practice.

Ask About the Rule

This hypothetical pair shows how to remove the client file from the research task.

Unsafe request: “Read the attached client loan ledger and tell me whether their repayments satisfy Division 7A.”

Safer request: “Explain how minimum yearly repayments work for an existing Division 7A loan. Link to Australian Taxation Office guidance and explain which loan details an accountant needs to assess.”

The safer request contains a general question. It leaves actual amounts and transactions with the practice. Removing a name alone leaves too much information if the remaining facts identify the client.

Before sending a question, remove client names, contact details and identifiers. Leave out actual figures and attachments. For calculations needed to understand a rule, use clearly fictional inputs that do not reproduce the client’s situation.

Keep privacy and answer accuracy separate. Open the cited guidance and read the relevant passage before applying an AI answer to a client file.

What Perplexity Keeps

Perplexity processes your questions and uploads to answer them. Switching off training changes a use of that information. The service still receives it.

Search History and Training Settings

The Perplexity Privacy Notice covers prompts, uploaded content and generated answers. It also describes account and device information. Its retention rule allows storage for service purposes and longer periods required or permitted by law.

Perplexity’s search and personal-data retention guidance says it keeps personal information while your account remains active. Treat an account’s history as stored research, including anything confidential you accidentally put into a question.

For a personal account, change the training setting before researching:

1. Open your account settings and go to Preferences.
2. Turn off AI data retention.
3. Confirm the toggle is off before sending the next question.

Perplexity’s data-collection guidance says the opt-out covers data collected afterwards. It cannot remove previously collected training data. Service operation and other stated processing can continue after you opt out.

Uploaded Files

Perplexity’s file-upload guidance gives session attachments a 30-day retention period, reduced to seven days for Enterprise Pro. Files in Projects and personal repositories remain until deleted under their applicable policies.

An expired attachment can still leave information in earlier answers. Removing a file from the next question also leaves previous answers with their existing context. Delete the session when you need its attached files removed.

A public session can expose its uploaded files to anyone with its link. Keep practice research private and share a public source directly when a colleague needs the underlying guidance.

Enterprise Data Terms

The Enterprise terms prohibit Perplexity from using, or authorising others to use, Customer Content to train or improve generative AI models. They also incorporate a Data Processing Addendum for personal data processed on the customer’s behalf.

These terms still permit processing needed to provide the service. A training restriction does not mean that client information stays on your device or that every stored copy disappears immediately.

Enterprise retention settings require at least 50 Enterprise Pro seats or one Enterprise Max seat. Session retention runs from the last activity, so a follow-up question restarts the period. Project files have separate retention rules.

For your practice’s approval, record the account type and permitted information alongside the chosen retention settings. Staff need to know which account they can use before they open a client matter.

The Comet Browser

Comet can send page content to Perplexity when the Assistant needs it to complete your request. That includes the current page and other tabs you ask it to use.

Perplexity’s Comet Assistant privacy guidance says the Assistant does not upload your full tab list by default. A page summary, selected-text request or multi-tab task can send the required content.

The guidance gives page and tab context a storage period of up to 30 days for history features. A Temporary Session can have its expiry removed. The broader Privacy Notice also describes browsing-history collection depending on your settings.

An open client portal can therefore become input to a research request. Keep client portals and practice email outside your Comet research session. A general question needs public guidance, so open the public source in its own research environment.

To restrict the Assistant, open Settings, then Privacy and security, then Comet Assistant. You can block it on specific websites or disable it entirely. For a practice with client portals, block those domains before staff start using Comet.

Test the restriction on a harmless page covered by the same website rule. The Assistant must be blocked there. Keep client work in your approved browser while you use Comet to research public material.