Rule · OSC · 12 Nov 2026
Registered firms where Ontario is the principal regulator — exempt market dealers, investment fund managers, portfolio managers, restricted dealers and restricted portfolio managers — must maintain policies and procedures under Section 11.1 of NI 31-103 addressing AI-specific risks, oversee third-party AI providers, and deal fairly, honestly and in good faith under Section 2.1 of OSC Rule 31-505, including in AI marketing claims. Current registrants must file a Form 33-109F5 if AI use changes their primary business activities, target market, or client products and services.
As part of the Registration, Inspections and Examinations (RIE) Division’s 2025 Examination Priorities, RIE committed to examining the prevalence of Artificial Intelligence (AI) systems within registrants’ operations and how registrants are responding to Canadian Securities Administrators (CSA) Staff Notice and Consultation 11-348 Applicability of Canadian Securities Laws and the use of Artificial Intelligence Systems in Capital Markets (CSA SN 11-348). Accordingly, RIE Staff (Staff) developed a multi-phase examination initiative (the OSC AI Initiative) with the following objectives:
To gain an understanding of the AI systems adopted by registrants and the manner in which those systems are being used; and
To assess registrants’ compliance with applicable securities law requirements when using AI systems, informed by the guidance provided under CSA SN 11-348.
The first phase of the OSC AI Initiative was completed in Fall 2025 with the development and issuance of the OSC AI Usage and Governance Survey for Registrants (the OSC AI Survey). The 30-question OSC AI Survey asked registrants to identify their current and intended AI system use and focused on obtaining a practical understanding of the adoption, deployment, governance, and oversight of such systems. The OSC AI Survey also sought details of any challenges or successes experienced by registrants in their use of AI systems.
The second phase of the OSC AI Initiative involved conducting focused examinations of a group of registered firms (the OSC AI Sweep). The examinations focused on observing how registrants are using AI systems in practice, with particular attention given to use cases identified in the OSC AI Survey that may directly affect clients, including, but not limited to, investment decision-making, portfolio management, and marketing. The examinations also included an assessment by Staff of the registrants’ system of controls supporting AI system use, including the established AI governance and oversight framework.
The OSC AI Initiative provided Staff an opportunity to gather relevant and timely data to support registrants in their adoption of AI systems, while enhancing the OSC’s understanding of how AI systems are being used by registrants and assessing the potential for investor harm. Staff recognize the rapid pace of global AI system adoption, and ongoing dialogue with registrants will help inform oversight and the responsible deployment of AI systems in Canadian capital markets.
The purpose of this Staff Notice is to share Staff’s observations from the OSC AI Initiative, including general information from the OSC AI Survey on the use of AI systems by registrants in Ontario. In addition, this Staff Notice provides a summary of the findings from the OSC AI Sweep and builds on existing guidance regarding the use of AI systems. Guidance provided in this Staff Notice is based on existing securities laws and does not create any new legal requirements, nor modify existing ones.
Staff issued the OSC AI Survey in November 2025 to all registered firms where Ontario is the principal regulator, encompassing a total of 787 firms registered as either exempt market dealers, investment fund managers, portfolio managers, restricted dealers, or restricted portfolio managers, or any combination of the above categories. Respondents represented a broad cross-section of registrants, with diverse business models and operational scales, ranging from small and medium-sized firms to large institutional registrants. Registrants were provided with a two-week response period to complete the OSC AI Survey. The OSC AI Survey achieved an overall response rate of 86%. Staff thank all registrants for their participation and cooperation in completing the OSC AI Survey.
The following observations are based on the OSC AI Survey responses received and should be interpreted in that context. The OSC AI Survey results reflect registrants' responses to the multiple-choice options provided and are intended to identify broad trends, rather than provide a complete assessment of any individual registrant's AI use, governance, or compliance practices.
Out of all responding registrants that use AI systems, the most commonly cited goal of their AI systems adoption is to improve efficiency (96%). Cost reduction and decision-making support were the next most commonly cited goals.
Out of all responding registrants that currently use or are exploring the use of AI systems, the most cited challenges relate to technical readiness, particularly integrating AI systems with existing systems (50%), data quality and accessibility (43%), and lack of internal expertise (42%).
Although the nature of challenges registrants face varies by registrant, the results suggest that technical readiness remains a key barrier to AI adoption and implementation.
96%
59%
58%
36%
28%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Improve efficiency
Reduce costs
Support decision-making
Enhance customer experience
Create/Augment/Innovate products and/or services % of Registrants Using AI Systems What are the goals of using AI systems in your firm?
50%
43%
42%
40%
34%
20%
18%
16%
13%
0%
10%
20%
30%
40%
50%
60%
Integration with existing systems
Data quality or ability to access data
Lack of internal expertise
Unclear return on investment
Regulatory compliance
Ethical concerns
Resistance to change
Difficulty explaining AI system outputs
Lack of funding
% of Registrants Using and/or Exploring AI Systems What are the challenges your firm faces with AI system adoption and implementation?
The OSC AI Survey set out a non-exhaustive list of broad functions commonly applicable to registrants’ operations and asked registrants to identify the areas in which they were using AI systems, and the extent of that use (i.e., minimal, limited, broad, integral, or not applicable).
Most common use cases observed related to administrative tasks, market research and analysis, sales and marketing, communications with clients, compliance-related tasks, and cybersecurity.
Overall, the responses suggest that AI adoption among many registrants remains at an early stage.
Across all reported use cases, associated AI systems use was most often characterized as minimal or limited, indicating that many registrants are piloting AI systems or using them in select areas rather than integrating them more broadly across business operations.
At the same time, some registrants reported broad or integral use of AI systems in certain functions, indicating more advanced deployment in specific areas, such as the most common use cases mentioned above.
The OSC AI Survey covered a non-exhaustive list of possible use cases, however, whether an AI system is used in a manner consistent with applicable securities legislation depends on the specific facts and circumstances and must be assessed on a case-by-case basis.
The OSC AI Survey responses also provided insight into how registrants using AI systems are governing and overseeing their use. Many registrants reported having governance processes in place, including approval processes, responsible use guidelines, formal policies and procedures, and human review of AI system inputs and outputs.
The responses indicated that governance and oversight practices varied across registrants. Registrants reporting limited or no formal AI governance framework or AI-specific policies were primarily those with limited AI system use and non-client facing use cases, suggesting that governance practices may still be evolving for firms whose use remains at an early stage.
The following charts summarize responses to the OSC AI Survey regarding:
components commonly included in registrants’ AI governance frameworks;
how registrants incorporate human oversight into the review of AI system outputs and validation of data accuracy; and
how registrants address data accuracy, security and privacy considerations relating to AI system use.
Registrants surveyed and currently using AI systems identified having implemented a range of AI governance framework components. Among that subset of registrants, established approval processes, ethics guidelines, and oversight roles are the most common governance components in place with:
o 73% having an approval process before adopting AI systems;
o 54% having AI ethics and responsible use guidelines; and
o 48% having defined roles and responsibilities for AI systems oversight.
However, nearly 20% of surveyed registrants currently using AI systems reported having no formal governance framework related to AI systems use.
What components of an AI governance framework does your firm currently have in place?
Approval process before the adoption of AI systems AI ethics and responsible use guidelines Defined roles and responsibilities for AI systems oversight AI strategy aligned with business objectives Governance committee or working group for AI systems oversight Regulatory compliance tracking Risk management protocols specific to the use of AI systems Internal audit or review mechanisms for AI systems Model documentation and version control Explainability and transparency standards No formal governance framework in place 73% 54% 48% 46% 44% 44% 44% 31% 31% 29% 19% 0% 10% 20% 30% 40% 50% 60% 70% 80% % of Registrants Using AI Systems
The presence of “human-in-the-loop” processes is one of the recommended AI Governance and Oversight policies and procedures mentioned in CSA SN 11-348 for effective monitoring.
While AI adoption among many registrants remains in the early stages, with common AI use cases to date being associated with internal business functions, registrants surveyed appear to be taking a cautious approach, with 82% always requiring human review and approval of AI system inputs and/or outputs before any action is taken.
In contrast, 5% require human approval specifically for high-risk systems only, 12% apply human-in-the-loop controls sometimes, depending on the use case, and 1% reported no human intervention.
Is there a “human-in-the-loop” process when using AI systems in your firm?
Yes, always reviews and approves
Yes, for high-risk systems only
Sometimes, depending on use case
No human intervention
82%
5%
12%
1%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
% of Registrants Using AI Systems
The majority of surveyed registrants currently using AI systems reported having implemented foundational data governance controls, with the most widely adopted data governance controls being access controls and user permissions (68%), compliance with relevant regulations (65%), and data encryption and secure storage protocols (61%).
Less than 30% of registrants conduct regular audits of data handling practices or inputs, or use third-party quality tools, and only 6% have no formal measures in place at all.
68%
65%
61%
47%
36%
29%
25%
18%
6%
0%
10%
20%
30%
40%
50%
60%
70%
Access controls and user permissions
Compliance with relevant regulations (e.g., PIPEDA) Data encryption and secure storage protocols Maintain documentation of data source and licensing Anonymization or pseudonymization of sensitive data Regular audits of data handling practices Regular audits/reviews of data inputs for quality Use of third-party data quality tools or platforms None - No formal measures in place % of Registrants Using AI Systems How does your firm address data accuracy, security and privacy concerns related to AI system use?
76%
75%
69%
58%
54%
50%
9%
0%
10%
20%
30%
40%
50%
60%
70%
80%
Staff training and awareness programs
Policy updates and procedural changes
Implementation of technical safeguards (e.g., access controls, monitoring) Escalation to risk committees or senior leadership Vendor or third-party reassessment Adjustments to AI system design or use cases None - No formal mitigation strategies in place % of Registrants Using AI Systems How are identified AI-related risks addressed or mitigated?
56% of surveyed registrants currently using AI systems indicated they provide no formal AI disclosure, while the remaining 44% provide disclosure in various forms and to different stakeholders.
Of those registrants providing disclosure, the most common delivery method was verbal explanation, followed by written disclosures and online disclosures.
Following the completion of the OSC AI Survey, Staff conducted focused examinations of a sample of 17 registrants as part of the OSC AI Sweep, which included firms registered in the categories of investment fund manager, portfolio manager, and exempt market dealer, and combinations thereof.
Staff’s examinations focused on better understanding how registrants were using AI systems in practice, the types of systems being used, and the accompanying controls and supervision in place to manage the business risks associated with its AI system use considering the obligations under Section 11.1 of National Instrument 31-103 Registration Requirements, Exemptions and Ongoing Registrant Obligations (NI 31-103). Staff evaluated whether registrants had addressed the following areas in managing the risks associated with AI system use:
AI system policies and procedures, covering the use of AI systems and handling of AI system risks
AI governance and oversight framework
Due diligence and ongoing oversight of third-party AI service providers
Disclosure to clients regarding the use of AI systems and services
In addition, Staff evaluated how registrants described or promoted their use of AI systems in public-facing or client-facing materials, including whether those statements were consistent with the requirement under subsection 2.1(1) of OSC Rule 31-505 Conditions of Registration (OSC Rule 31-505), for registered dealers and advisers to deal fairly, honestly and in good faith with their clients.
Deficiencies were raised if Staff noted a concern in how business risks pertaining to AI system use were being managed or where disclosure concerns were identified based on the nature of the AI use case and the impact on firm clients.
This Staff Notice includes examples of control elements registrants had in place to manage business risks associated with AI system use, while maintaining compliance with applicable securities law requirements. Staff strongly recommend that registrants review CSA SN 11-348 in its entirety, as this Staff Notice builds on that guidance.
Under Section 11.1 of NI 31-103, registered firms must maintain policies and procedures that establish a system of controls and supervision that provide reasonable assurance that the firm, and individuals acting on its behalf, comply with securities legislation and manage the risks associated with its business in accordance with prudent business practices.
Staff recognize that AI use cases present varying levels of risk and that higher risk uses, particularly those affecting investor outcomes, require stronger supervision. The extent of human oversight should be proportionate to the nature and complexity of an AI system, and the potential impact of AI system use on clients or the registered firm.
Staff identified several registrants that had adopted AI systems but had not updated their policies and procedures to address the unique features and risks associated with AI use.
Policies and procedures should address AI-specific features and risks and may be incorporated either as a standalone AI policy or within existing policies. When developing AI-related policies and procedures, registrants should address the following elements:
The purpose and scope of AI system use;
Governance and approval process for AI systems and use cases;
Data governance and privacy standards;
Oversight of AI systems, including human involvement;
Conflicts of interest;
Client disclosure and communication, including marketing considerations; and
Training.
In addition, Staff observed that some registrants lacked structured employee training to support the implementation of their AI-related policies and procedures. The role of training is to supplement the policies and procedures to reinforce the registrant’s responsibility in managing the associated risks of using AI systems.
By contrast, over the course of the OSC AI Sweep, Staff observed certain good practices in respect of training, including:
Setting a minimum or frequent cadence for AI-specific training and reinforcing that training to all staff within the registrant is an important part of overall compliance oversight
Frequent communication to all staff within the registrant to reinforce or update key AI use requirements, in alignment with the registrant’s policies and procedures
Tracking training completion and adoption rates within the registrant’s business
Staff identified registrants who relied on an AI system developed by a third-party (as a third-party service provider) but maintained limited or no evidence of the due diligence activities performed. These include both the initial due diligence performed before first engaging the service provider, as well as ongoing monitoring and supervision.
Staff also noted that some registrants maintained limited documentation demonstrating that they:
sample the output and verify the accuracy of third-party AI systems on an ongoing basis;
oversee, review, and control the use of and ongoing configuration of the third-party AI systems used; and
identify and manage the associated risks with third-party AI system use.
Under Part 11 of the Companion Policy to NI 31-103, registrants are responsible and accountable for all functions that they outsource to a service provider, must undertake due diligence before contracting for outsourced services, and must supervise any outsourced service provider on an ongoing basis. Registrants will need to develop tailored policies and procedures to address the specific risks posed by using AI systems developed and operated by third parties.
Staff also remind registrants of the risks associated with providing third-party AI systems with access to the registrant’s systems or data, as doing so may expose confidential information, including personal client details. Part of any initial and ongoing due diligence of the registrant’s third-party service providers should be considered from a cybersecurity perspective, which was broadly highlighted as part of CSA Staff Notice 33-322 Review of Registered Firms’ Cybersecurity Practices and Additional Guidance.
Registrants are also required to maintain such documentation of their due diligence pertaining to a third-party service provider as part of their books and records under Section 11.5 [General requirements for records] of NI 31-103.
Staff note that strong investor interest in AI systems can lead to practices where market participants make inaccurate, false, misleading, or embellished claims to attract investors to capitalize on the growing interest in the use of AI. Referred to as “AI washing”, these claims are designed to imply that a market participant has a competitive advantage based on the use of AI systems. Such claims can mislead clients and influence investment decisions based on incomplete or unsupported information.
Examples of inaccurate, false, misleading, or embellished claims Staff identified over the course of the OSC AI Sweep included:
Claims of the prediction accuracy of the AI system or tool for certain types of market events, where the actual number of predictions was lower than stated for that type of market event.
Statements exaggerating the AI system or tool being “best of” or as having a competitive advantage, which were not substantiated.
With respect to disclosure of AI use, Staff observed instances where registrants clearly explained how AI systems are used in their business. Staff observed examples of relationship disclosure information and other client-facing disclosure describing AI system use that included information relating to:
the purpose of AI use and the tools used for those use cases;
how AI outputs are used in those use cases;
how client data is used and protected;
the risks associated with the use cases; and
how the registrant oversees the use of AI systems, including any human-in-the-loop controls.
Registrants should assess whether their use of AI systems triggers disclosure obligations to clients under applicable securities legislation, including the client relationship disclosure requirements in NI 31-103, and the obligation to deal fairly, honestly and in good faith with clients. This assessment should include whether AI is sufficiently integrated into registrable services such that a reasonable investor would consider it important to the client’s relationship with the firm. Any disclosure should be clear, meaningful, and proportionate to the AI use case and related business risks.
Registrants must deal fairly, honestly and in good faith with their clients under Section 2.1 of OSC Rule 31-505, which extends to any claims or messages provided in marketing materials or disclosures to the client.
Staff also identified instances in which benchmarks were used to highlight the performance of the AI systems; however, in some cases, no disclosure was provided to clarify how those benchmarks were calculated or determined. As with the case of any claims regarding the performance or use of an AI system, registrants should ensure any statements and comparative information provided are fairly presented and disclosed, consistent with the registrant’s obligation to deal fairly, honestly and in good faith under Section 2.1 of OSC Rule 31-505.
Staff expect registrants to apply a proportionate, risk-based approach when adopting AI systems. The nature and extent of the controls, governance, and oversight applied to an AI system should be commensurate with the nature, scale, and complexity of the risks presented by the particular use case.
When developing AI-related policies and procedures, registrants should consider covering the following areas:
o Governance & accountability – e.g., clear ownership and senior oversight, model/data risk management, responsibility over AI system use and training
o Lifecycle controls – e.g., risk-based controls over design, testing, deployment, and ongoing monitoring
o Human oversight – e.g., effective human-in-the-loop processes that are scalable and proportionate to the nature, purpose, scope, and risks of the AI use case, allowing registrants to determine the appropriate level of human monitoring and intervention
o Explainability and record-keeping – e.g., analyzing input data and AI system outputs, managing for AI system bias, and hallucinations
o Outsourcing and third-party risk – e.g., due diligence, ongoing vendor oversight and supervision, and privacy law implications
o Disclosure – e.g., informing clients and prospective clients of the registrant’s AI use, associated risks, and any conflicts of interest, without misrepresentation
For third-party AI systems, registrants must evaluate and document their understanding of the system before adoption and continue to evaluate and monitor it during use. This may include the following practices:
o when evaluating an AI system, registrants could gain an understanding of the risks of using the third-party AI system in each of the following risk areas:
explainability, including model architecture
model bias
cybersecurity and data privacy
client impact
o continuous oversight and monitoring may be evidenced through logs or periodic reviews of the performance of the third-party AI system used, such as identifying instances of anomalies, model drift, or errors.
When promoting the use of AI systems, ensure that any statements or claims regarding the AI system and its use are substantiated and fairly represent the performance of the AI system in the registrant’s operations.
In addition, and as described in CSA SN 11-348, Staff remind registrants applying for registration that they must describe their business plans, including operating models, in their applications. The use of AI systems in ways that may directly affect registerable services provided to clients must be disclosed in these applications.
Current registrants must file a Form 33-109F5 Change of Registration Information (F5 Filing) having regard to the requirements of National Instrument 33-109 Registration Information, if they change their primary business activities, target market, or the products and services they provide to clients.
o Staff acknowledge that the determination as to whether a F5 Filing is required before adopting a specific AI system must be based on the specific facts and circumstances supporting the AI use case, which will differ on a case-by-case basis.
o In instances where a registrant wants to discuss whether the use of AI systems in their operations requires a F5 Filing, they are encouraged to contact their legal or compliance advisors and/or OSC Staff directly. Any discussions will assist in determining whether an F5 Filing is required.
The results of the OSC AI Survey show that most registered firms have either adopted the use of AI systems or plan to in the near future. OSC Staff continue to support the responsible adoption and use of AI systems in capital markets where investors and market participants can benefit from their use while ensuring that the associated risks are appropriately mitigated.
OSC Staff are assessing the use of AI systems in capital markets, including in the provision of investment advice and governance practices by market participants. This assessment is being conducted in collaboration with other jurisdictions in Canada through the CSA’s Artificial Intelligence Working Group. Staff encourage registrants and other market participants to participate in future consultations to support a comprehensive regulatory understanding of this technology and any regulatory actions that may be required to support its responsible adoption in capital markets.
Registrants are encouraged to use this Staff Notice, together with CSA SN 11-348, as a self-assessment tool to assess compliance with their regulatory obligations. Staff will continue to monitor the use of AI systems by registrants through the compliance examination process. Where instances of non-compliance with securities legislation are identified, we will take appropriate action.
If you have any questions regarding this Staff Notice, please refer them to any of the following:
Daniel Panici
Head of Portfolio Management Compliance
Registration, Inspections and Examinations dpanici@osc.ca
Rachel Keane
Senior Legal Counsel
Registration, Inspections and Examinations rkeane@osc.ca
Carlin Fung
Senior Accountant
Registration, Inspections and Examinations cfung@osc.ca
Vincent Chow
Senior Accountant
Registration, Inspections and Examinations vchow@osc.ca
April Zhang
Accountant
Registration, Inspections and Examinations yzhang@osc.ca
Note from RegAlert. AI assistants can read this document in full, and search 70,000+ more, through the RegAlert MCP connector (https://mcp.regalert.today/mcp). Free with an account. How to connect ChatGPT, Claude or Cursor.
More like this from OSC
OSC published 5 documents in the last 30 days. We email you each new one the day it's published.
Source: Ontario Securities Commission — original document · Summary generated with machine assistance and checked by an AI reviewer before publication; the authoritative text is the regulator's original document. How RegAlert works
In RegAlert: added