Ontario’s Evolving Approach to Trustworthy AI

The interview was conducted by Ula Rutkowska (Senior Researcher, Apolitical) and edited by Christina Obolenskaya (MSc in International History, LSE and Communications Intern, Apolitical). Stephanie Oldfield was named on Apolitical’s Government AI 100 2025.

Ontario’s government has been taking a proactive approach to artificial intelligence (AI), setting requirements for its responsible use through the Trustworthy AI Framework. This evolving framework ensures AI is deployed in a way that aligns with public values, mitigates risks, and enhances government services. With the recent passage of the Enhancing Digital Security and Trust Act, 2024, Ontario now has a legal foundation to regulate AI use in the public sector.

In this Q&A, we speak with Stephanie Oldfield, Director of the Digital and Data Policy Branch, in the Ministry of Public and Business Service Delivery and Procurement, about the challenges and opportunities of implementing AI in government. From overcoming hesitancy around AI adoption to fostering a culture of appropriate experimentation, Stephanie shares insights into how Ontario is navigating this complex and fast-moving policy landscape. We also explore the role of data governance, real-world AI applications, like AI Scribe, and what it takes to build an AI-ready public sector.

Most importantly, we discuss the future of AI in Ontario and how Ontario’s approach will continue to evolve.

Ontario's Trustworthy AI Framework sets rules for the responsible use of AI in government. What do you see as the most critical components of this framework and how have they been implemented so far?

The trustworthy AI framework is constantly evolving. It’s an excellent example of truly iterative policy development, starting in 2021 when we heard from individuals that they wanted clarity from government about how and where AI would be used.

We got a lot of feedback from individuals all along the way, starting with feedback on principles for AI use in Ontario. It has continued to grow over the past few years alongside the growth in AI use, and in January, 2025, the Enhancing Digital Security and Trust Act came into force in Ontario, which defines artificial intelligence systems for the province and allows us to develop regulations about how AI is used in the public sector.

The most important thing about the framework is that we started early (AI years seem to feel like dog years — every one year feels like seven) and that we’re not aiming for a one-and-done policy approach. It’s our job to protect you, to support you and the economy, and to help Ontario succeed, which is a big responsibility. With the trustworthy AI framework, being comfortable with its existence in a way that will continue evolving as the technology evolves is critical. The most critical component about the framework isn’t a specific component itself, but that we haven’t paused on AI policy development. It will never be complete, as there is no definitive right time with things like AI. Things are moving very fast, and the flexibility of the framework means we can respond quickly and thoughtfully. So, while you might have reviewed principles for AI use as part of the framework two years ago, now you can review the province’s Directive for Responsible AI Use — and there may be new things in the future, too.

What would you say have been some of the biggest challenges in implementing the Trustworthy AI Framework across government operations and what have you done to address them?

There's certainly a hesitancy or reticence sometimes to just try new things, which can be a big barrier. Trying new things can be hard and frustrating, whether it’s a hobby or a sport or an AI tool. As soon as someone tries to use an AI tool once or twice, though, it's demystified, and folks have much more clarity about how and where they might integrate AI into their work. Encouraging people to try using it for the first time, even for small things, is such an important hurdle to get past. We do a lot of demos and walk people through different uses to help give them a feel for it.

It also helps to show the areas where they might not think there would be any risk. We developed guidance and best practices for using Generative AI (GenAI) and showcasing the intersection of things like personal information and GenAI tools is extremely important. Something as innocuous as making a colleague a birthday card could result in someone’s biometric information being shared without their consent. This isn’t unique to the public sector or the nature of our work — but it’s something we’re all learning together as these tools become more accessible and widespread. Folks hear a lot about the risks of AI but unless we couple that conversation with practical tools and advice about how to manage those risks, we aren’t going to move the needle on appropriate AI use.

Another drawback of AI moving so fast is that everyone feels like they are late to the conversation. I like to remind folks that it’s OK for them to feel like they don’t fully understand everything about AI; AI experts in academia and huge private sector businesses running AI models don’t, either! The important thing is to be thoughtful, be curious, and apply critical thinking. Giving people space and grace to ask any question, without feeling like they should have somehow earned a PhD in machine learning on the side of their desk this past year, is one of the biggest ways we can meaningfully increase AI literacy together. This applies across the board, not just in the public service — with your parents, your friends, your communities.

How does Ontario's approach to data governance support the principles of trustworthy AI? What improvements are you prioritizing?

We have extremely smart folks thinking specifically about data and how we need to modernize our data governance to reflect the new realities, including the use of AI systems. A big part of the conversation has to be with people, though. What do they think is appropriate?

To me, that’s a critical part of where our data conversation needs to evolve to, and it’s really an emphasis on the “conversation” part. How can we, as governments, get better at articulating the kinds of benefits strong data governance can bring to people and to their lives? What are people comfortable with? Where do they see those boundaries for governments? We can never have that conversation too often, as people’s thinking is always evolving, and it’s not often front-of-mind like it can be for other organizations that use our data where we interact with them more frequently, like mobile apps.

It’s a complex conversation, but one we need to trust the public to have as we consider the potential of AI over the coming months and years — and that’s something that’s top of mind for me.

__Can you think of any particular ways in which AI is being used to improve government services in Ontario? __

One example I can share is AI Scribe implemented through the Ministry of Health. My dad actually got an email from his physician asking if he consented to its use in his appointment, and I often use my dad as a benchmark for people who don't know anything about digital or tech (love you, dad!). It described in a plain-language way how AI would be used to summarize his conversation so that his physician could spend more time on his care. I thought for sure he would decline, but he consented and said it was logical. This is a really person-facing way that AI is being used in government and we're starting to see it more and more.

What could other organizational leaders or team leads be doing to foster and facilitate an AI-ready culture in their team or their organisation?

Every single day, every conversation I go into, I'm expecting to learn something that I didn't know before about AI.

We really have to encourage and facilitate an environment where everybody, no matter how senior or junior, is motivated to try things and to learn things. Where they feel comfortable and confident to raise issues and bring forward ideas, flags, whatever the case might be.

What is your vision for the role of AI in Ontario's government operations, and how do you see the trustworthy AI framework evolving in the future?

Because AI is changing so much, I don’t think it’s useful to have a specific vision of an end state where we're using AI in specific ways, or in specific amounts. It depends on how fast the technology moves, on what barriers and risks come up, and how we manage them. If we get focused on those kinds of outcomes it will be easy to lose our way. I want to be focused on making sure that we understand what people want and need, so that strong AI use can flourish and can help make the lives of people that live here better.

I am on the policy side of AI, but there's also an important tech side and tech uptake side of things, of course. We foster an environment where public servants (and the public) don’t think this is only a technology conversation.

I hope we continue to talk and work with people who live in this province about what they think is appropriate and what they're worried about, having a dialogue between government, the public, and experts. AI is overwhelming, and it often feels like the future being painted for us is either utopian (AI gives us so much time back!) or dystopian (AI takes over humanity). There are so many potential outcomes in between those two, and we all have a role in building the kind of AI-enabled future we want.