Imagine you are trying to build a robot that can help make important decisions, like who gets a visa, which medical scan needs a doctor's immediate attention, or who gets hired for a job. In the world of computer science, this is called Artificial Intelligence (AI) governance. Think of AI governance as the rulebook and the safety harness for these robots. Without it, a robot might make a mistake that hurts someone, or it might be biased against certain groups of people.

For a long time, experts have had a few different rulebooks. One famous one, from the United States, is called the NIST AI Risk Management Framework. It's like a very detailed, high-quality instruction manual on how to build safe and fair robots. It tells you to "Govern" (have a boss), "Map" (know your risks), "Measure" (test your work), and "Manage" (fix problems). Another set of rules comes from the European Union, which is more like a strict law that says "you must do this or you get fined."

But here is the tricky part: Canada is a bit like a giant house with many different rooms, and each room has its own set of rules. The federal government (the "house owner") has one set of rules for its own robots. But the provinces (the "rooms," like Ontario, Quebec, and Alberta) have their own, slightly different rules. Some rooms care mostly about privacy, others care about who is in charge, and some care about how the robot explains its choices. For a company or government trying to build a robot that works in all these rooms, following every single rulebook separately is like trying to wear five different pairs of shoes at once. It's confusing, expensive, and you might trip over your own feet.

This is where the paper comes in. The authors, led by Mohammad Etemad from Scitus Solutions, are proposing a new way to solve this shoe problem. They created a framework called SCITUS (Systematic Canadian Integration for Trustworthy and Unified Standards). Think of SCITUS as a "universal adapter" or a "master key." Instead of trying to follow five different rulebooks, SCITUS takes the popular NIST instruction manual and tweaks it just enough to fit perfectly into every Canadian room at the same time.

The paper argues that the old way of doing things—checking the rules for Ontario, then stopping to check the rules for Quebec, then checking the federal rules separately—is a waste of time and money. The authors suggest that by using their new SCITUS framework, organizations can build one set of safety controls that satisfies everyone. They tested this idea with three made-up stories: a federal immigration robot, a hospital robot in Ontario, and a hiring robot in Quebec. In each story, they showed how SCITUS could help the robot pass all the tests without needing to build five different versions of itself.

The paper doesn't claim to have solved every possible problem in the world, nor does it say that the rules will never change. Instead, it suggests that this "adapter" approach is a much smarter, faster, and safer way to handle the messy reality of having different laws in different places. It's a guide for how to be a good robot builder in Canada, ensuring that the robots we trust are fair, safe, and ready for whatever room they walk into.

---

The Paper in Plain English: The "Universal Adapter" for AI Rules

The Problem: The "Five Pairs of Shoes" Dilemma
Imagine you are a robot builder in Canada. You want to build a smart system to help sort through thousands of visa applications. But you have a huge headache: Canada is like a house with many different rooms, and each room has its own strict rulebook for how your robot must behave.

• The Federal Government (the main house) has a rulebook called the Treasury Board Directive. It says, "If your robot makes big decisions, you must fill out a specific form called an Algorithmic Impact Assessment (AIA) and prove you checked for bias."
• Ontario (one of the rooms) has a new law, Bill 194. It says, "If you are a public building, you must publish a list of your robots and explain how you keep them safe."
• Quebec (another room) has Law 25. It says, "If your robot uses personal info, you must tell the person immediately and let them ask a human to review the decision."
• Alberta has its own rules about accuracy and privacy.
• British Columbia has some general advice but no strict laws yet.

If you try to follow all these rules separately, it's a nightmare. You'd have to write five different reports, run five different tests, and hire five different teams of lawyers. The paper says this is like trying to wear five different pairs of shoes at the same time. It's heavy, confusing, and you're likely to trip.

The Solution: The SCITUS "Universal Adapter"
The authors created a new tool called SCITUS. Think of it as a "universal adapter" for your robot's safety gear. Instead of buying five different adapters for five different outlets, SCITUS is one smart adapter that fits them all.

It works by taking a famous, high-quality rulebook from the US called the NIST AI Risk Management Framework (which is already used by many companies worldwide) and "translating" it for Canada.

• The Core: It keeps the four main steps from NIST: GOVERN (have a boss), MAP (find the risks), MEASURE (test the robot), and MANAGE (fix the problems).
• The Twist: It adds "Canadian flavor" to each step. For example, when the NIST manual says "be transparent," SCITUS says, "Be transparent and make sure you tell people in French and English, and tell them exactly what Quebec's Law 25 requires."

How It Works: The Three-Layer Cake
The paper describes SCITUS as a three-layer cake:

1. The Bottom Layer (The Foundation): This is the original NIST framework. It's the sturdy base that keeps everything aligned with international standards.
2. The Middle Layer (The Canadian Mix): This is where the magic happens. The authors took all the rules from the federal government and the provinces (Ontario, Quebec, Alberta, etc.) and mixed them into the NIST steps. They figured out which rules overlap (so you only have to do them once) and which rules are unique (so you don't miss them).
3. The Top Layer (The Frosting): This is the practical guide. It gives you checklists, templates, and tools so you can actually build the robot without getting lost.

The "Control Catalog": A Growing Toolbox
The paper mentions that they have a "Control Catalog" (a list of safety rules) that is growing.

• In June 2025, they had 31 rules.
• By July 2026 (the time of this paper), they updated it to 57 rules.
  Why did it grow? Because the world of AI is changing fast. New threats appeared, like "agentic AI" (robots that can act on their own) and new rules about how training data is collected. The authors added new rules to handle these fresh challenges, showing that their system can adapt as the world changes.

Testing the Adapter: Three Stories
To prove their idea works, the authors told three stories (scenarios) about how a company would use SCITUS:

1. The Immigration Robot (Federal):

   • The Challenge: A robot that sorts visa applications. It has to be fair, bilingual, and secure.
   • The Conflict: The government wants to be transparent about how the robot works, but security experts say, "If we tell everyone exactly how it works, bad guys will trick it!"
   • The SCITUS Fix: The framework helped them find a middle ground. They published general info (like "we check for fairness") but kept the secret sauce (the exact math) private. They also solved a tricky math problem: how to be fair to all countries when some countries have more qualified applicants than others. They chose a specific definition of "fairness" (called equalized odds) and documented why they chose it.

2. The Hospital Robot (Ontario):

   • The Challenge: A robot in a Toronto hospital that looks at X-rays to find emergencies. It has to follow Ontario's public sector rules, Health Canada's medical device rules, and privacy laws.
   • The Conflict: Usually, a hospital would have to fill out one form for the province, another for the federal health agency, and another for privacy.
   • The SCITUS Fix: The framework showed that these rules actually overlap a lot. By using SCITUS, the hospital could write one safety plan that satisfied all three regulators at once. They saved weeks of work and made sure the robot was safe for patients.

3. The Hiring Robot (Quebec):

   • The Challenge: A private company in Quebec using AI to hire employees. They have to follow Quebec's strict privacy laws (Law 25).
   • The Conflict: The company needs to test if the robot is biased against certain groups, but privacy laws say they can't collect data about people's race or ethnicity.
   • The SCITUS Fix: The framework suggested a clever workaround. They could ask people to voluntarily share their background for testing purposes (without using it for the hiring decision), or use "proxy" data (like the city someone lives in) to guess if the robot is fair. This let them test for bias without breaking privacy laws.

What the Paper Says (and Doesn't Say)
The authors are careful to say that SCITUS is a suggestion and a framework, not a magic wand that fixes everything instantly.

• They suggest that using this "universal adapter" is much more efficient than trying to follow every rulebook separately. They estimate that following rules one by one takes 12 to 16 weeks per system, while using SCITUS could cut that down to 6 to 8 weeks.
• They rule out the idea that you can just pick the "easiest" rule to follow. They say you have to follow the strictest rule in the room to be safe.
• They don't claim that this solves every legal problem. They admit that sometimes rules conflict (like transparency vs. security), and in those cases, you need a human committee to make a tough choice. SCITUS just gives that committee a better map to make the choice.

Why It Matters
The paper concludes that for any country with different rules in different places (like the US or Australia), this "adapter" approach is a smart way forward. It stops organizations from wasting money on duplicate work and helps them build AI that is actually safe and fair for everyone, no matter which "room" they are in.

In short, SCITUS is the guidebook that says: "You don't need five different pairs of shoes. You just need one really good pair that fits every room in the house."