Short answer: no. There is no public database anyone can search to see which AI systems are operating in Canada, who runs them, or what they do. Two things exist that are sometimes mistaken for one, and a genuine example elsewhere shows exactly what a real AI registry would look like.
Key takeaways
Searching for “Canada AI registry” turns up two real things and a lot of confusion about what each one actually is. Neither is a registry of AI systems in the sense most people mean the word — a searchable list of who is running what.
The Treasury Board’s Directive on Automated Decision-Making requires a department to complete, approve and publish “the final results of an algorithmic impact assessment in an accessible format on the Open Government Portal prior to the production of any automated decision system” (TBS, Directive on Automated Decision-Making, s.6.1.1). That does put real information about real federal automated systems into a public, government-run location — which is closer to a registry than anything else Canada has. But look at what it actually covers before calling it one: only systems that fall under this specific directive, which applies to systems “used to make an administrative decision or a related assessment about a client” inside the federal government. A provincial government system, a municipal one, or a private company’s AI product is not in it, and never will be under this instrument — it was never built to cover them.
ISED’s Voluntary Code of Conduct publishes a named list of 46 signatory organizations — including TELUS, Cohere, CIBC and the Organisme d’autoréglementation du courtage immobilier du Québec (OACIQ) — each committed to a measures table covering accountability, safety, fairness, transparency, oversight and robustness (ISED, Voluntary Code signatories). That list answers “which organizations have committed to this code,” and it is a real, public, named list. It does not answer “which AI systems exist and what do they do” — a signatory’s name on the list says nothing about which specific products it operates, how many, or where. It is a registry of commitments, not of technology, and the code itself is explicit that signing changes nothing about “existing legal obligations that organizations may have — for example, under” PIPEDA. Non-signatories, meanwhile, are simply absent from the list entirely, whether or not they operate AI systems in Canada — the list only ever shows who opted in.
The contrast is sharper against a jurisdiction that actually built a registry. The European Union’s AI Act creates a genuine legal obligation: providers of certain high-risk AI systems are required to register themselves, and information about the system, in an EU database established and managed by the European Commission — and public-authority deployers must register before they use such a system (Regulation (EU) 2024/1689, recital 131 — European Union; consolidated text, the official version being published on EUR-Lex). That is what a real AI system registry looks like: a specific legal duty, tied to a specific category of system, enforced by requiring registration as a precondition of lawfully placing the product on the market. Nothing in the Canadian material sourced here comes close to that shape — not the TBS impact assessments (federal government only, not a precondition of building the system, just a publication duty), and not ISED’s signatory list (voluntary, company-level, not system-level).
Right now, nothing in Canada. There is no form to check, no database to search, and no legal duty requiring a private business anywhere in the country to register an AI system with any government body before deploying it. A federal government department is the closest exception, and even there the obligation is to publish an impact assessment, not to “register” a system in the sense the EU’s database does. If Canada were to build a genuine AI registry — something Bill C-27’s proposed Artificial Intelligence and Data Act did not itself create either, since it left the definition of a “high-impact system” to future regulations rather than fixing a registration duty in the bill's own text — the EU’s database shows one concrete model of what that would need to include: a specific trigger (a defined risk category), a specific obligation (register before market), and a specific consequence for not doing it.
Suppose a Canadian business wants to know whether a competitor’s AI-driven product has been reviewed, registered or disclosed to any government body before it went to market. Today, that search has nowhere to go. There is no federal registry to query — the Open Government Portal’s algorithmic impact assessments cover only the government’s own internal systems, not products a private company sells. There is no provincial equivalent sourced here either. Checking ISED’s signatory list would only tell you whether the competitor’s organization voluntarily signed the code — and a competitor with a perfectly ordinary, unremarkable AI product has exactly as much reason to sign as one with a genuinely risky one, so presence or absence on that list says very little about the product itself. The honest answer to give a client asking this question today is that the check does not exist yet in Canada, not that the answer came back clean.
Bill C-27’s proposed Artificial Intelligence and Data Act is the clearest evidence that a registry was not the government’s chosen mechanism even when it tried to legislate this space. As introduced at first reading, the bill would have required a person responsible for a high-impact system to, “in the time and manner that may be prescribed by regulation, publish on a publicly available website a plain-language description of the system” (Bill C-27, first reading, cl.11(1)) — each provider publishing its own disclosure on its own site — rather than filing into one central government-run database the way the EU’s regulation requires. That is a meaningfully different design choice: decentralized, per-provider disclosure instead of a searchable central list. Whether that would have functioned as a practical substitute for a registry is a separate question from whether AIDA, even if it had passed as introduced, would have created one — and on the bill’s own text, it would not have.
Related: what AIDA proposed and where it stands and how public bodies in Canada disclose AI use.
No. There is no public Canadian database of AI systems in use by private businesses to search, and none is created by any statute currently in force.
It is a public list ISED publishes, but it registers organizations that voluntarily committed to a code of conduct — not AI systems, and not a legal requirement to appear on it.
No. It applies only to federal government automated decision systems under the Treasury Board’s own directive — a province, municipality or private business is outside its scope entirely.
A short conversation can map what applies today and what a future registry-style rule might ask for.
