Summary#
The Public Sector Artificial Intelligence and Cybersecurity Governance Act creates a new law; it does not amend another Act. Its stated goal is to set rules for public-sector use of artificial intelligence (AI) and cybersecurity. The Act starts only on a date set by proclamation, and no start date is given in the supplied material.
- Regulations will decide which public-sector entities are covered. The Act lists government, certain organizations defined under The Financial Administration Act, municipalities, northern incorporated communities and local government districts.
- For covered entities, regulations may require public information about AI use, accountability frameworks, risk-management steps and rules for how AI may be used. Regulations may also prohibit particular uses.
- Covered entities must follow prescribed AI technical standards and cybersecurity requirements and standards. In prescribed circumstances, an individual must oversee AI use and provide additional information, as regulations require.
- The minister may issue a cybersecurity directive to a specific covered entity, with approval from the Lieutenant Governor in Council. The entity must comply, and the directive must be made public.
- The law allows different rules for different entities or classes. Regulations generally require public consultation, unless the minister considers the matter urgent.
What it means for you#
- Public-sector entities: If regulations prescribe your organization, it must follow the applicable AI and cybersecurity rules. The Act does not itself set out the detailed rules; those are left largely to regulations.
- People using public services: The Act does not itself give you a right to notice, an explanation, a human decision-maker, or a way to challenge an AI-supported decision. Regulations may require some public information and individual oversight.
- Universities, health authorities and school divisions: The explanatory note gives these as examples, but the Act’s definition does not name them. Whether any are covered depends on whether they fit its defined categories and what regulations prescribe.
- General public: The Act does not itself ban a particular AI use or require every public-sector entity to disclose its AI use. Those rules may be set by regulation.
Money#
No cost information is in the available material.
- The Act requires covered entities to meet prescribed requirements, but gives no estimate for staff, systems, training or other compliance costs.
- It does not set an appropriation, fee, fine or benefit.
What is unclear#
- Which entities and AI systems will be covered, and when the Act will come into force, depend on future regulations or proclamation.
- The Act does not set the detailed standards, prohibited uses, reporting rules or oversight procedures.
- It does not create a complaint or appeal process, or say that non-compliance invalidates a policy or decision. In fact, it says failure to comply does not affect the validity of listed policies, instruments or decisions.
- The Act sets no specific penalties or enforcement powers for non-compliance.
- It is unclear how these rules will interact with existing programs and protections. The Act says another enactment prevails if it conflicts with this Act.
Case for#
- A possible argument for the Act is that it creates a way to set common rules for public-sector AI and cybersecurity, while allowing different requirements for different kinds of entities.
- Regulations could require records, risk assessments, bias testing, incident reporting and human oversight. These tools could help make technology use more accountable, if regulations require them.
- Public consultation and later reviews of regulations could give the public a role in shaping the rules and assessing how well they work.
Case against#
- Much of the practical protection depends on future regulations. The Act itself does not say which AI uses are prohibited or what standards entities must meet.
- The Act does not provide a direct complaint or appeal process, and non-compliance does not invalidate decisions. People affected by an AI-supported decision may therefore lack a remedy under this Act.
- The minister can set technical standards by regulation, and can issue different cybersecurity directives to specific entities. The Act gives little detail about how those standards or directives will be developed beyond the consultation rule for regulations.