AI Governance8 min read

Ontario Has Three AI Governance Frameworks. Most Organizations Know About One.

Ontario's OPS AI Directive, IPC-OHRC Principles, and Bill 149 each shape AI governance. Here is what each requires and what to do about it.

What You'll Learn

How Ontario's three AI governance frameworks (OPS Directive, IPC-OHRC Principles, Bill 149) align and point the same way, the four operational components every organization needs in place, and why the window between published principles and enforced requirements is when compliance costs the least.

Ontario's three-layer AI governance system consists of the OPS AI Directive (governing provincial ministries and agencies since December 2024), the IPC-OHRC Joint Principles for Responsible AI Use (principles written for Ontario's public and broader public sector since January 2026, and a useful lens for private organizations), and Bill 149's AI hiring disclosure requirement (binding legislation for employers with 25+ employees since January 2026). Together they show where Ontario's regulators and government are heading on AI.

Ontario has quietly built a three-part AI governance picture since 2024. Bill 149 received Royal Assent in March 2024 and its AI hiring disclosure took effect in January 2026. The OPS AI Directive arrived in December 2024. The IPC-OHRC Principles followed in January 2026. Each was released separately by a different branch of government. Most organizations read about one, skimmed another, and missed the third entirely. That is a problem, because the three pieces point the same way, and the IPC and OHRC have published a comparison showing the Principles and the Directive align and complement one another.

The three layers are the Responsible Use of Artificial Intelligence Directive (the OPS AI Directive), the IPC-OHRC Joint Principles for Responsible AI Use, and Bill 149's AI disclosure requirements for hiring. Taken individually, each looks like a manageable policy update. Taken together, they show where Ontario is heading, and they give any Ontario organization using AI in operations, hiring, or customer-facing systems a practical benchmark.

Here is what each framework requires and why the combined effect matters more than any single piece.

The OPS AI Directive: The Government's Own Rulebook

The Responsible Use of Artificial Intelligence Directive took effect December 1, 2024, as a Management Board of Cabinet Directive under the Management Board of Cabinet Act. Ontario.ca

It applies to all Ontario ministries and provincial agencies and sets requirements across the full AI lifecycle: design, development, procurement, deployment, operation, and decommissioning.

Three requirements stand out for private sector organizations watching regulatory direction:

First, centralized AI governance. The Directive mandates a consistent, centralized approach to AI governance across all government entities. This signals that Ontario regulators view ad hoc, department-by-department AI adoption as a risk, not a feature.

Second, mandatory risk assessment. Before deploying AI, ministries must complete AI risk assessments under the Directive, alongside existing privacy and threat-risk processes. Ontario.ca

Third, disclosure and reporting. Ministries must disclose and report on AI use. Ontario publishes its AI use cases in a public registry. Ontario Open Data

The Directive applies to government. But government procurement contracts flow downstream. Any vendor selling AI-enabled services to an Ontario ministry or agency will need to demonstrate that their systems meet the Directive's risk management standards. The compliance requirement travels through the supply chain.

Not sure where AI fits in your operations?

Take the Free AI Readiness Scorecard →

The IPC-OHRC Principles: The Evaluation Framework for Everyone

On January 21, 2026, the Information and Privacy Commissioner of Ontario (IPC) and the Ontario Human Rights Commission (OHRC) released six joint principles for the responsible use of AI. IPC Ontario

The IPC-OHRC Principles are written for Ontario's public sector and broader public sector institutions. Private organizations are not bound by them, but they are the regulators' stated lens on responsible AI use, so they are worth treating as a benchmark. The six principles are: valid and reliable, safe, privacy-protective, human-rights affirming, transparent, and accountable.

What makes the Principles strategically significant: the IPC and OHRC published a comparison showing the Principles and the OPS AI Directive align and complement one another. The comparison appears in a letter to the Secretary of Cabinet published alongside the Principles. IPC Ontario

The alignment is intentional. The government set its own standards first, and the privacy and human rights regulators then published principles that align with them. The Principles are not legislation. They carry no fines today. But they establish the evaluation criteria that regulators will use when assessing complaints, conducting investigations, or advising on future legislation.

We published a full analysis of the IPC-OHRC Principles and their operational implications in our guide to the IPC and OHRC AI evaluation framework.

Bill 149: The Hiring Disclosure Mandate

Bill 149, the Working for Workers Four Act, introduced Ontario's first binding AI disclosure requirement for employers. Effective January 1, 2026, any employer with 25 or more employees using AI to screen, assess, or select candidates in a publicly advertised job posting must disclose that fact in the posting itself. Ontario Legislature

Bill 149 is narrow in scope. It covers hiring AI, not operational AI. But it established a precedent: Ontario is willing to legislate AI disclosure requirements at the activity level. The pattern is disclosure first, then broader regulation.

For organizations with AI in multiple business functions, Bill 149 is the signal that hiring was the starting point, not the finish line. Our detailed breakdown covers the Bill 149 compliance requirements.

The Pattern: Regulators Building Toward Full AI Governance

Read the three frameworks together and the regulatory strategy becomes clear:

💡

Ontario's AI governance now has three parts: internal government standards (OPS Directive, Dec 2024), principles for the public and broader public sector (IPC-OHRC Principles, Jan 2026), and enforceable legislation for one business function (Bill 149, in effect Jan 2026). The IPC and OHRC's comparison shows the Principles and the Directive align and complement one another (IPC Ontario).

Layer 1 (OPS Directive, Dec 2024): Ontario governs its own AI use with lifecycle risk management, impact assessments, and public disclosure.

Layer 2 (IPC-OHRC Principles, Jan 2026): Ontario's privacy and human rights regulators publish principles for public sector institutions that align with the government's framework.

Layer 3 (Bill 149, Jan 2026): Ontario legislates its first binding AI disclosure requirement for a specific business function. We published a detailed guide on what employers must know about the AI hiring disclosure requirement.

The trajectory runs from internal government standards to published principles to enforceable legislation. The IPC-OHRC letter comparing the Principles to the OPS Directive shows the two align and complement each other. If you need an operational starting point, begin with an AI readiness assessment that addresses compliance.

For comparison: the federal Artificial Intelligence and Data Act (AIDA) died when Parliament was prorogued in January 2025 and has not been reintroduced. Ontario is not waiting. It is building provincial AI governance through a series of smaller, interlocking moves that collectively create a compliance framework no single piece of federal legislation has achieved.

The Gap Between Adoption and Governance

The urgency is practical, not theoretical. 88% of organizations report regular AI use in at least one business function, per McKinsey's 2025 State of AI survey (McKinsey). Yet most organizations see disappointing returns because employees experiment with AI tools without integrating them into actual workflows. Harvard Business Review

A separate March 2026 HBR analysis identified the "last mile problem" as the primary obstacle: organizational design, legacy processes, and governance gaps stall AI adoption at the employee level, even when the technology works. Harvard Business Review

📊
Example

Consider a hypothetical Ontario professional services firm that adopts AI across four functions: hiring (ATS with resume scoring), client onboarding (chatbot), financial reporting (anomaly detection), and internal scheduling (calendar optimization). Each tool was procured by a different department. No central inventory exists. Under Bill 149, the HR team does not know the ATS qualifies. When a client files a privacy complaint about the onboarding chatbot, no one can produce the documentation the IPC-OHRC Principles require. The governance gap is invisible until enforcement arrives.

✅
Result

Organizations that build governance proactively — AI inventory, risk assessments, disclosure protocols, and accountability structures — address all three frameworks simultaneously. The cost of building governance before enforcement is a fraction of the cost of building it during an investigation.

56% of the CEOs surveyed said they are getting "nothing out of" their AI investments, per PwC's 29th Global CEO Survey shared at the World Economic Forum in January 2026.

These numbers describe the exact gap Ontario's three frameworks address. Organizations adopted AI before building governance around it. The frameworks now require governance to catch up to adoption.

What Operational Compliance Looks Like

An organization that takes Ontario's three-layer framework seriously needs four things in place:

An AI inventory. You cannot govern what you have not catalogued. Every AI system in use across operations, hiring, customer service, and decision-making needs to be documented. The IPC-OHRC Principles' accountability and transparency principles point to the same practice.

Risk assessment by system. Each AI system needs a documented risk assessment covering privacy impact, human rights impact, and algorithmic bias. The OPS Directive requires AI risk management for ministries, and it is easier to build in at the design stage than after deployment.

A disclosure protocol. Bill 149 already requires disclosure in hiring. The IPC-OHRC Principles expect transparency across all AI use. Organizations should build a disclosure framework that scales beyond hiring to any customer-facing or employee-facing AI system.

An accountability structure. Someone in the organization owns AI governance. Not as a side project attached to the CTO or legal counsel. As a defined responsibility with authority to pause deployment, require assessment, and report to leadership.

The Window

Ontario's AI governance frameworks are in their early enforcement period. Bill 149 is binding. The IPC-OHRC Principles carry regulatory weight in investigations. The OPS Directive governs procurement. Organizations that build governance now operate ahead of enforcement. Organizations that wait will be reacting to regulatory inquiries instead of demonstrating compliance proactively.

The gap between AI adoption and AI governance is where regulatory risk accumulates. Ontario has told organizations what it expects. The question is whether your organization can demonstrate compliance when asked.

💡
Key Takeaways
  • Ontario's three AI governance frameworks (OPS Directive, IPC-OHRC Principles, Bill 149) align — the IPC and OHRC's letter to the Secretary of Cabinet compares the Principles and the Directive — and together they set a practical benchmark across hiring, operations, and customer-facing AI.
  • The four operational requirements are an AI inventory, risk assessments per system, a scalable disclosure protocol, and a named accountability owner with authority to pause deployments.
  • The trajectory is clear: internal government standards first, external evaluation criteria second, enforceable legislation third. Building governance now costs a fraction of building it during a regulatory inquiry.

If you are evaluating where your organization stands, book a free discovery call and we will map your AI use against Ontario's regulatory expectations.

Or take the free AI Readiness Scorecard.

Frequently Asked Questions

What are Ontario's three AI governance frameworks?
Ontario has three overlapping AI governance frameworks: the Responsible Use of Artificial Intelligence Directive (OPS AI Directive, effective December 2024) governing all Ontario ministries and agencies, the IPC-OHRC Joint Principles for Responsible AI Use (January 2026), written for the public and broader public sector, and Bill 149's AI disclosure requirement for publicly advertised job postings (effective January 2026, employers with 25 or more employees). Together they show where Ontario is heading on AI.
Does the OPS AI Directive apply to private sector organizations?
The OPS AI Directive directly applies to Ontario ministries and provincial agencies. However, private sector organizations are affected through government procurement requirements, as vendors selling AI-enabled services to Ontario government entities must demonstrate their systems meet the Directive's risk management standards. The IPC-OHRC Principles were written for the public and broader public sector.
What is the connection between the OPS AI Directive and IPC-OHRC Principles?
The IPC and OHRC published a comparison showing the IPC-OHRC Principles and the OPS AI Directive align and complement one another, in a letter to the Secretary of Cabinet published alongside the Principles. The government set its own AI governance standards first, and the privacy and human rights regulators then published principles that align with them.
What does an organization need for Ontario AI compliance?
Organizations need four components: an AI inventory documenting every AI system in use, risk assessments for each system covering privacy, human rights, and algorithmic bias, a disclosure protocol that covers hiring (Bill 149) and scales to other AI uses, and an accountability structure with a defined owner for AI governance who has authority to pause deployment and require assessments.