Skip to content

Territory 5 · AI governance

Can Canadian Financial Advisors Use ChatGPT? A Compliance Analysis

PIPEDA, CIRO and Quebec Law 25 are three regulatory layers that US guidance does not cover. What each ChatGPT tier means for Canadian advisors under them.

Updated 14 min readSandra Lyne
An instrument maker’s hands setting a small offset on a brass dial.

Canadian financial advisors can use ChatGPT for general research, drafting marketing content, and exploring concepts. But the moment client personal information enters a prompt, three Canadian regulatory frameworks apply that US-focused guidance does not address: PIPEDA, CIRO's supervision and record-keeping requirements, and Quebec's Law 25. The answer depends on which version of ChatGPT you use and what you type into it.

In 2024-2025, 59% of the business advisory files at the Office of the Privacy Commissioner of Canada involved advice on adopting AI, up from 40% the year before (2024-2025 Annual Report to Parliament (external site, opens in a new tab), 5 June 2025). The question is not hypothetical.

Advisors describe using these tools for meeting notes and summaries, client email drafts, and marketing content. Jason Pereira of Woodgate Financial described to Advisor.ca in 2023 how he uses an AI meeting assistant that produces an interactive transcript of a client call (Advisor.ca, 28 September 2023 (external site, opens in a new tab)).

A Broadridge survey (external site, opens in a new tab) fielded in late 2023 found that 56% of the 131 Canadian advisors surveyed were using or planning to use generative AI in digital marketing, against 43% of US advisors. Statistics Canada reported (external site, opens in a new tab) that 30.6% of finance and insurance businesses had used AI to produce goods or deliver services in the 12 months to the second quarter of 2025, up from 10.9% a year earlier.

The advisors asking "can I use ChatGPT?" are not asking a theoretical question. They are asking about a tool they already use on their phones between meetings. The compliance analysis that follows is written for the Canadian regulatory environment, because much of what is published online is written for US advisors under SEC and FINRA rules that do not apply here. If you are evaluating Claude instead, see the companion Claude compliance analysis.

Which version of ChatGPT are you using?

This is the question that does most of the compliance work.

OpenAI offers several tiers of ChatGPT, and they handle your data differently. The compliance implications are not the same.

Free, Go, Plus and Pro (plans for individuals): As at 13 November 2026, OpenAI states that it may use content (external site, opens in a new tab) from its services for individuals to train its models unless the user opts out. The opt-out has existed since April 2023, and the default is that training is on. OpenAI lists its business commitments (external site, opens in a new tab) (no training on the organization's data by default, and a Data Processing Addendum) for ChatGPT Business, Enterprise and Edu and for its API platform. The plans for individuals are not on that list, so no data processing agreement stands between an advisor's firm and OpenAI, and no Canadian data residency is offered.

Under PIPEDA's framework, this may amount to a "disclosure" of personal information to OpenAI for its own purposes, not a "transfer for processing." That distinction is significant. In its Home Depot findings (external site, opens in a new tab) (PIPEDA Findings #2023-001, 26 January 2023) the Office of the Privacy Commissioner concluded that Home Depot should have obtained express opt-in consent before disclosing e-receipt customers' information to Meta, which could use it for its own business purposes, because customers would not reasonably expect the practice. Typing client details into an individual ChatGPT plan without the client's express consent for that use raises the same question.

Business (formerly Team): As at 13 November 2026, OpenAI states that it does not train on a Business workspace's data (external site, opens in a new tab) and supports a Data Processing Addendum for business customers. Data residency is listed for eligible Enterprise, Edu, Healthcare and API customers, and Business is not among them.

Enterprise: The plan with the most contractual and administrative controls. As at 13 November 2026, OpenAI states that it does not train its models on an organization's data by default, supports a Data Processing Addendum, and provides audit logs, admin controls and single sign-on. OpenAI lists Canada among the regions where eligible new Enterprise and Edu workspaces can store customer content at rest (external site, opens in a new tab). Storage is not the same as processing. OpenAI offers inference residency (keeping the model's computation in region) in Europe, the United States and the United Arab Emirates, and does not list Canada. A workspace with Canadian data residency therefore has its content stored in Canada and has no commitment from OpenAI that the model's computation happens there. Even the most configured Enterprise deployment carries residual exposure under the US CLOUD Act, which allows US legal process to compel data from US companies regardless of where that data is physically stored.

As at 13 November 2026, OpenAI asks firms to contact its sales team (external site, opens in a new tab) for the Enterprise plan. It is a plan a dealer or a larger firm arranges, not one an advisor signs up for alone.

The practical upshot: An individual plan (Free, Go, Plus or Pro) comes with no data processing agreement between your firm and OpenAI, and PIPEDA's accountability principle calls for "contractual or other means" to provide a comparable level of protection when a third party processes personal information. Anything that could identify a client stays out of those plans. Whether a Business or Enterprise workspace meets a firm's requirements is a decision for the firm's compliance officer, and where the model's computation happens is one of the things to weigh.

One common misconception: using the ChatGPT desktop app (available on macOS and Windows) does not alter any of this. The desktop app is an interface. Every prompt you type is still sent to OpenAI's cloud infrastructure for processing. "On your computer" and "processed on your computer" are not the same thing. The same PIPEDA obligations, the same cross-border transfer considerations, and the same record-keeping limits apply regardless of whether you access ChatGPT through a browser, a phone, or the desktop application.

Three Canadian regulatory layers ChatGPT triggers

PIPEDA: you remain accountable for data sent to OpenAI

PIPEDA does not prohibit cross-border data transfers. The OPC's 2009 Guidelines for Processing Personal Data Across Borders (external site, opens in a new tab) take an accountability-based approach: "The transferring organization is accountable for the information in the hands of the organization to which it has been transferred." Principle 4.1.3 requires "contractual or other means" to provide "a comparable level of protection."

The same guidelines say an organization must weigh all the elements of a transfer, and add: "The result may well be that some transfers are unwise because of the uncertain nature of the foreign regime or that in some cases information is so sensitive that it should not be sent to any foreign jurisdiction." Northern Catalyst's view is that detailed KYC documentation, health disclosures and estate planning details, the kind of information that flows through a typical advisor-client meeting, are the information that sentence asks an advisor to think hardest about.

On 6 May 2026 the Privacy Commissioner of Canada, with the privacy regulators of Quebec, British Columbia and Alberta, published the findings of a joint investigation of OpenAI's ChatGPT (PIPEDA Findings #2026-002 (external site, opens in a new tab)). The complaint was well founded, including a finding that OpenAI had not obtained valid consent for its collection, use and disclosure of personal information in developing and deploying its GPT-3.5 and GPT-4 models. Under PIPEDA it was "conditionally resolved" on the strength of commitments OpenAI made, with deadlines of three and six months from the report and quarterly reports to the regulators until all are met. For advisors, the relevant point is that Canadian privacy regulators have applied existing privacy law to this specific tool.

The December 2023 joint principles (external site, opens in a new tab) signed by all 14 federal, provincial, and territorial privacy commissioners added another layer: AI-generated inferences about identifiable individuals are treated as a new collection of personal information. If you ask ChatGPT to analyze a client situation, the inferences it generates create a fresh privacy obligation you may not have considered.

CIRO: existing supervision rules already cover AI output

As at 13 November 2026, CIRO has published no rule or guidance note specific to AI. That does not mean existing rules are silent.

CSA Staff Notice and Consultation 11-348 (external site, opens in a new tab) (5 December 2024) sets out how existing securities law applies when registrants use AI systems. It is staff guidance and, in its own words, "does not create any new legal requirements, nor modify existing ones." The existing duties stay with the registrant. If ChatGPT hallucinates a product recommendation or fabricates a regulatory citation, the advisor who sends it to a client owns it. CIRO's Compliance Report for 2026 (external site, opens in a new tab) (Administrative Bulletin 26-0034, released 17 February 2026) says: "As part of our FinOps examination approach, we will be enquiring about the use of AI in dealers' operations. To the extent dealers are using AI, we will be reviewing the operational controls they implemented to ensure AI is working as designed."

ChatGPT conversations also sit outside the firm's records. A registered firm must keep its required records for seven years from the date the record is created (NI 31-103, s. 11.6(1)). A personal ChatGPT account gives the firm no retention control: the user can delete a conversation, and the firm has no hold over it.

Quebec Law 25: a categorically different regime

Advisors operating in Quebec or serving Quebec-resident clients face additional requirements that go beyond PIPEDA. Law 25 (section 17, in force since 22 September 2023) requires a privacy impact assessment before personal information is communicated outside Quebec. The Act draws the line at Quebec's border, not Canada's, so it reads as covering a transfer to another province as well as to another country. The assessment takes into account the sensitivity of the information, the purposes for which it is to be used, the protection measures that would apply, and the legal framework of the place receiving it. The information may be communicated if the assessment establishes that it would receive "adequate protection."

Law 25 sets two tiers of penalty for an enterprise: administrative monetary penalties of up to $10 million or 2% of worldwide turnover, whichever is greater, and penal fines of up to $25 million or 4% of worldwide turnover, whichever is greater. The Act also provides for civil claims: where an unlawful infringement of a right under the Act causes an injury and is intentional or results from a gross fault, the court awards punitive damages of at least $1,000 (section 93.1).

How Canadian rules differ from US guidance

If you have searched for guidance on using ChatGPT as a financial advisor, most of what you found was written for the US market. The regulatory frameworks are materially different.

RequirementUS (SEC/FINRA)Canada (CIRO/PIPEDA)
Data privacy lawRegulation S-PPIPEDA + Quebec Law 25
Record retentionSEC Rule 204-2 (5 years)IDPC Rule 3800; the seven-year keeping duty sits with the registered firm under NI 31-103 s. 11.6(1)
SupervisionFINRA Rule 3110CIRO supervision rules (no rule or guidance note specific to AI, as at 13 November 2026)
Client consent for AI processingNo explicit federal requirementPIPEDA Principle 4.3 (meaningful consent)
Cross-border data transferNo federal restrictionPIPEDA cross-border accountability obligations
Advisor accountability for AI outputSEC fiduciary duty / FINRA suitabilityExisting securities law applies when registrants use AI systems (CSA Staff Notice 11-348)
Regulators' notices on AIFINRA Regulatory Notice 24-09 (27 June 2024)CSA Staff Notice and Consultation 11-348 (5 December 2024)

The differences are not minor. Canadian registered firms keep their required records for seven years. Canadian privacy regulators have published findings on the specific tool in question. And under PIPEDA's accountability principle the organization that transfers client data, not OpenAI, remains accountable for it after it crosses the border.

Guidance written for US advisors answers a different set of rules.

What has happened so far

Several published sources show how these tools behave in practice.

An academic study (external site, opens in a new tab) first posted in November 2024 tested 150 citations and found that ChatGPT-4o had a 20% hallucination rate when asked for references in the financial literature.

In Moffatt v. Air Canada, 2024 BCCRT 149 (external site, opens in a new tab), British Columbia's Civil Resolution Tribunal found Air Canada liable for negligent misrepresentation after a chatbot on its website gave a customer inaccurate information about bereavement fares. Air Canada had argued the chatbot was "a separate legal entity that is responsible for its own actions." The tribunal member wrote: "This is a remarkable submission." and "It should be obvious to Air Canada that it is responsible for all the information on its website." It is a small claims decision, and the award was $650.88 in damages. Its reasoning turns on a company's responsibility for the information it gives customers, whatever tool produced it, which is the position an advisor is in when sending a client content a tool drafted.

A survey of 500 Canadian accountants and bookkeepers (external site, opens in a new tab), commissioned by Dext and fielded by Censuswide in December 2025, found that 76% had seen an increase in clients using public AI tools for financial, tax or bookkeeping advice, and 50% were aware of businesses that had suffered direct financial losses from it. The same pattern appears in adjacent Canadian professions.

None of these involved regulatory discipline against a Canadian advisor. The 2026 Compliance Report means dealers can expect to be asked about their use of AI and the operational controls around it.

What ChatGPT is well suited to

ChatGPT has legitimate uses for advisors where no client information is involved.

The decision rule is straightforward: does the input contain any information that could identify a client? If no, the privacy obligations described above are not engaged. PIPEDA's obligations are activated by the presence of "personal information" in the input.

These uses include drafting general marketing copy, researching financial concepts, creating templates and checklists, writing practice descriptions, summarizing publicly available regulatory documents, and brainstorming content ideas. None of these involve personal information, so the privacy questions above do not arise on any plan.

For client-specific work (meeting notes, suitability analysis, follow-up correspondence), the question becomes how to get the value of AI processing inside the rules described above. Jason Pereira's advice to advisors in the same article is direct: "Do not ever put client information into ChatGPT or any open protocol" (Advisor.ca, 28 September 2023 (external site, opens in a new tab)). Some advisors try to meet that standard by removing identifying details by hand before they paste text in. The challenge is that manual redaction is error-prone. Advisors forget details, abbreviate inconsistently, or miss secondary identifiers.

Large firms have faced the same question. JPMorgan Chase and Goldman Sachs were reported in 2023 to have restricted or blocked staff use of ChatGPT (Semafor, 19 May 2023 (external site, opens in a new tab)). Morgan Stanley built an internal assistant with OpenAI and reported in June 2024 (external site, opens in a new tab) that 98% of its financial advisor teams had adopted it.

The question is not whether advisors will use AI. It is whether they will use it through channels that meet their regulatory obligations.

What to do now

The 2026 Compliance Report says CIRO's examiners will be enquiring about the use of AI in dealers' operations. It adds that dealers "should also assess whether any use of AI and/or automation of its regulatory functions is a material business change". Preparing ahead of that conversation is straightforward.

Three steps are within every advisor's control.

First, document what AI tools you currently use and what data flows through them. If that includes client personal information through personal ChatGPT accounts, move that work to a tool your firm has approved before your next compliance review.

Second, talk to your compliance officer. If your firm has not published guidance on AI use, raising the question early is a constructive step. The free Compliance Conversation Kit includes a preparation guide and a meeting agenda for that conversation.

Third, for client-specific work, use a tool your firm has approved, under an agreement that covers where client information is processed, how long it is kept, and who can see it. The four questions in Getting Started with AI as a Canadian Financial Advisor apply to any tool.

At the end of the path

Compliance Conversation Kit

Free. The material for the conversation with the compliance officer.

Written by

Sandra Lyne

Founder, Northern Catalyst

TagsPIPEDACIROChatGPTAI ComplianceQuebec Law 25Financial Advisors

This is educational content based on published regulatory sources as at the page's last-updated date. It is not legal, compliance, or financial advice. Review specific compliance decisions with your compliance officer and qualified legal counsel before implementation.

Source: Joint CSA/CIRO Staff Notice 31-368 · 10 December 2025

Re-verified