The short answer: as of August 2026, no Canadian law requires you to tell a customer they are talking to AI rather than a person. That is the narrow answer, and it is the wrong one to act on. Three sets of rules that already bind a British Columbia business — the province's consumer protection statute, federal deceptive marketing law, and the CRTC's rules on automated calls — all attach to the impression your bot creates, not to the technology. And a BC tribunal has already held a company liable for what its website chatbot told a customer. Disclosure is not currently a duty. It is the cheapest way to avoid the thing that is.
The question comes up on almost every AI receptionist or website chatbot project now, usually within the first fifteen minutes: am I allowed to do this without saying it is a robot?
"Disclosure" here means a plain statement, early in the conversation, that the customer is dealing with an automated system rather than staff — not a footnote in your privacy policy, but a line the customer reads before they decide to trust the answer.
This is not legal advice. It is the practical version I give clients before we switch anything on. Anything unusual, ask a lawyer.
Is there a Canadian law that says you must?
No. There is no federal or British Columbia statute in force that names the duty and says "tell them it is a machine."
The reason people think otherwise is that a great deal has been proposed in 2026, and coverage of proposals reads like coverage of law. Bill C-36, the Protecting Privacy and Consumer Data Act, was tabled on 15 June 2026 and would add transparency duties around automated decision systems, with penalties up to C$10 million or 3% of global revenue. It has not passed. Bill C-34, the Safe Social Media Act, would regulate chatbot services, but it is aimed at large consumer platforms and child safety, not at your booking bot.
And on 23 July 2026 the federal government opened a public consultation on AI transparency that asks, in as many words, whether businesses should have to disclose when a customer service reply comes from AI rather than a person. Governments do not consult on rules that already exist.
What actually binds a BC business today
The absence of an AI-specific rule does not mean the absence of a rule. Here is the whole picture, sorted by whether it reaches a business operating in Kamloops.
| Rule | Binds a BC business today? | What it actually requires |
|---|---|---|
| BPCPA s. 4–5 (BC consumer protection) | Yes | No conduct with the capability or tendency to mislead a consumer in a consumer transaction |
| Competition Act, deceptive marketing | Yes | No false or misleading representations made to promote a product, service or business interest |
| Negligent misrepresentation (common law) | Yes | Reasonable care that what you tell a customer is accurate — including through a bot |
| BC PIPA | Yes | Meaningful consent and a stated purpose at or before you collect personal information |
| CASL | Yes, for commercial messages | Consent, sender identification and a working unsubscribe, whoever or whatever wrote the text |
| CRTC Unsolicited Telecommunications Rules | Yes, for outbound automated calls | Express consent before a synthesized-voice solicitation call; identify the caller |
| Bill C-36 (PPCDA) | Not in force | Would add automated decision system transparency |
| Bill C-34 (Safe Social Media Act) | Not in force | Would regulate chatbot services, child safety framing |
| EU AI Act, Article 50 | Only if you serve the EU market | Inform the user they are interacting with AI, no later than the first interaction |
| Utah AI Policy Act | Only if you serve Utah | Disclose generative AI on a clear consumer request; broader duty in regulated occupations |
Read down the "yes" rows and a pattern shows up. None of them mentions AI. All of them are about the impression a customer is left with. That is why "there is no disclosure law" is a true sentence that gets people into trouble.
The BC one deserves a closer look. Section 4(1) of the Business Practices and Consumer Protection Act defines a deceptive act or practice as a representation or conduct by a supplier that has "the capability, tendency or effect of deceiving or misleading a consumer." The customer does not have to prove they were actually fooled — the tendency is enough. And under section 5(2), once a deceptive act is alleged, the burden of proving it was not deceptive sits with the business, not the customer.
So the question is not "did I break the AI rule." It is "would a reasonable customer, reading this conversation, have believed they were talking to a person, and did anything turn on that." If your bot is called Sarah, signs off with a first name, and quotes a price, you are further along that road than you think.
The BC case that answers the liability question
The one people should actually know is Moffatt v. Air Canada, 2024 BCCRT 149, decided on 14 February 2024.
A man booking a flight after a death in the family asked Air Canada's website chatbot about bereavement fares. The bot told him he could apply for the reduced fare retroactively. Air Canada's actual policy did not allow that, and the airline argued the chatbot was a separate entity responsible for its own answers.
The tribunal rejected that. The chatbot is part of the company's website, so the company is responsible for what it says, in the same way it is responsible for a static page. Mr Moffatt was awarded $650.88 in damages for negligent misrepresentation, plus interest and fees.
Two things make this the most useful fact on the subject for a small operator in the BC Interior. It is a British Columbia decision. And it came from the Civil Resolution Tribunal — the online small-claims body that handles claims up to $5,000, needs no lawyer, and costs a filing fee. That is not a regulator you will probably never meet. That is the venue an irritated customer with a screenshot uses on a Tuesday evening.
The realistic downside of an unlabelled bot is not a federal penalty. It is one customer, one screenshot, and a quote your bot invented.
The one place there is already a hard rule: outbound AI calls
Website chat, inbound AI phone answering, and outbound AI voice calls are three different exposures, and almost every article on this topic treats them as one.
Inbound is the mild one. A customer rings your number and an AI answers. They initiated the contact, nothing was solicited, and the CRTC's telemarketing rules do not attach.
Outbound is where the hard rule lives. The CRTC's Unsolicited Telecommunications Rules define an automatic dialing-announcing device as automatic equipment able to store or produce phone numbers, used to convey "a pre-recorded or synthesized voice message." A solicitation call made with one requires express consent obtained before the call — not implied consent, not an existing business relationship. Telemarketers must also identify the caller, even on otherwise exempt calls.
An AI voice agent dialling a list of old quotes is a synthesized voice conveying a message to a phone number. Whether a two-way AI conversation fits a definition written for recorded announcements is genuinely unsettled, and the CRTC has opened a consultation — Compliance and Enforcement Notice of Consultation CRTC 2026-132 — that asks whether its own definition captures software, AI and other non-human generated voices.
Unsettled is not the same as safe. If you are considering outbound AI calling in Canada, treat the express-consent requirement as live and get advice first. The same instinct applies to automated texting, which sits under the consent rules that govern texting customers in Canada.
Does the EU AI Act reach a Kamloops business?
Almost certainly not, and it is worth saying plainly because the scare is everywhere right now.
Article 50 of the EU AI Act became applicable on 2 August 2026. It requires providers of systems intended to interact directly with people to ensure a reasonably informed user knows they are dealing with an AI system, no later than the first interaction, unless that is obvious from the context. It is the clearest disclosure rule in the world right now, and the model everyone else is drafting against.
It reaches you only if you place an AI system on the EU market or the system's output is used in the EU. A plumber in Kamloops with a booking bot does neither. If you sell software or services into Europe, that is a different conversation and a real one.
Same logic for the US state laws. Utah requires a business to disclose generative AI when a consumer clearly asks, with a broader duty for regulated occupations, since 7 May 2025. That binds you if you serve Utah customers. It is a preview, not an obligation.
The consultation you can actually answer
Canada's AI transparency consultation runs from 23 July to 23 September 2026 and is open to businesses, not just lawyers and lobby groups. Whether small operators have to bolt an AI banner onto every website chat window is one of the questions on the table, and the number of Kamloops-sized businesses that will submit a view is close to zero.
If you have an opinion about what disclosure should cost a four-person company, this is the month to file it.
How to disclose without killing the conversation
Five steps, and the whole thing takes an afternoon.
- Open with it, in the bot's first message. One clause is enough: "Hi — you're chatting with an automated assistant." Message one, before any question is answered, is the standard every proposed rule converges on.
- Do not give it a human first name. A bot called "Sam" with no label is the exact fact pattern that creates a misleading impression. Call it what it is: the booking assistant, the after-hours line, the AI assistant.
- Give a visible route to a person. "Ask for Jagat any time" or "reply HUMAN and I'll pass this to the owner." This is the part that protects the customer relationship, and it is also the part that converts, because the people who ask for a human are usually the ones ready to buy.
- Constrain what it can assert. Moffatt was a liability case about a wrong answer, not about a hidden identity. Ground the bot in your real pricing, hours and policies, and have it decline anything outside them rather than improvise. On this site, the chat assistant answers only from a fixed knowledge base and cannot invent a page link.
- Say what you keep. If the conversation is stored or emailed to you, BC's Personal Information Protection Act wants the purpose stated at or before collection. One line in the chat window and a matching line in your privacy policy covers it. The same care applies to what not to put into ChatGPT under BC privacy law.
None of that is expensive. Steps one to three are copy changes. Step four is how a bot should have been built anyway, and it is the difference between an AI receptionist that is worth the cost and one that generates refund arguments.
Why disclose when nothing makes you
Because the trend line is one-directional and the cost of getting ahead of it is a sentence.
The Pew Research Center found in June 2026 that 76% of US adults consider it extremely or very important to know whether pictures, videos or text were made by AI, while only 12% feel confident they could tell. That is American data, but it is the same direction the Canadian consultation is pointed.
The business case is simpler than the legal one. A customer who is told up front and gets a fast answer is pleased. A customer who works it out later feels handled. There is no third outcome, and only one of them leaves a review.
Common questions
Do you have to tell customers they are talking to AI in Canada? No Canadian law in force as of August 2026 requires it. But BC's Business Practices and Consumer Protection Act prohibits conduct with the capability or tendency to mislead a consumer, so an unlabelled bot that leaves a customer believing they spoke to staff can still create liability.
Is my business liable for what my chatbot tells a customer? Yes. In Moffatt v. Air Canada, 2024 BCCRT 149, the BC Civil Resolution Tribunal held that a chatbot is part of the company's website and the company is responsible for its answers. It rejected the argument that the chatbot was a separate entity, and awarded $650.88 in damages.
Does the EU AI Act apply to a Canadian small business? Only where the business places an AI system on the EU market or the system's output is used in the EU. Article 50's chatbot transparency duty became applicable on 2 August 2026, but a local service business with Canadian customers is outside its scope.
Can I use an AI voice to make outbound calls to past customers? Not without care. The CRTC's Unsolicited Telecommunications Rules require express consent before a solicitation call conveying a pre-recorded or synthesized voice message, and the CRTC is consulting in 2026 on whether that definition covers AI voices. Inbound AI answering is a different and much milder case.
What is the minimum disclosure that covers me? A single clause in the bot's first message stating that it is an automated assistant, a named route to a human, and a line about what is stored. That combination removes the misleading impression, satisfies BC PIPA's notice expectation, and takes an afternoon to implement.
The honest summary
Nothing in Canadian law currently forces you to label your bot. Label it anyway, in message one, and give people a way to reach you. The rule that will eventually exist is being consulted on right now, the rules that already exist are about the impression rather than the technology, and the tribunal that would hear the complaint is an online form your customer can fill in for the price of a filing fee. A clause of text is a cheap way to be finished with all three.
If you want the bot built so it discloses properly, answers only from your real pricing, and hands off to you cleanly, that is what the AI receptionist setup for Kamloops businesses covers, and how a small-business AI chatbot is actually set up walks through the build.
Sources: Have your say: advancing AI transparency in Canada (ISED) · Moffatt v. Air Canada, 2024 BCCRT 149 (CanLII) · CRTC Unsolicited Telecommunications Rules · Business Practices and Consumer Protection Act (BC Laws) · Pew Research Center, What do Americans think AI is? (17 June 2026)
Written by Jagatjeet — Jagatjeet (jagatjeet.com) is a web design, local SEO and AI automation studio in Kamloops, British Columbia, serving the Thompson-Okanagan and BC Interior. Published 21 August 2026. Last updated 21 August 2026. General information about Canadian and British Columbia rules, current at the date of publication — not legal advice.