The principles center on developing, deploying, and using AI such that public trust is maintained
The Information and Privacy Commissioner of Ontario and the Ontario Human Rights Commission have published joint principles on responsible artificial intelligence adoption.
The principles focus on helping organizations to maintain public trust by respecting privacy and human rights in developing, implementing, and using AI. Organizations using or are considering the adoption of AI systems are urged to adhere to six principles: valid and reliable, safe, protect privacy, affirm human rights, transparent, and accountable.
AI systems must generate valid, reliable, and accurate results for the purposes for which they were developed, used, or deployed. Systems must meet independent testing standards and objectively fulfill intended requirements for specific application. They must perform consistently over a certain duration in the environment in which they are implemented.
System use must comply with Ontario’s human rights and privacy laws, including the right to non-discrimination. AI systems must support considerations of human life, physical and mental health, economic security, and the environment; they must be reviewed and tracked to ensure their ability to withstand events or deliberate efforts to induce harm.
Thus, organizations must show that strong cyber security protection is in place along with human rights and privacy safeguards. New use of systems must pass a comprehensive assessment process to ensure safety.
AI system developers, providers, and users must proactively guard personal information and respect the right of access to information. Systems must be developed under a privacy by design approach that expects and minimizes privacy risks to groups and individuals, incorporating privacy protections into systems.
AI developers, providers, and institutions must ensure that systems do not infringe on significant equality rights by identifying inherent systemic discrimination and addressing them under the Ontario Human Rights Code. Governments and governmental actors must also respect the Canadian Charter of Rights and Freedoms, including the rights of freedom of expression, peaceful assembly, and association.
Organizations must be transparent about the use of AI systems and implement policies and practices that explain how systems work. Systems must be visible and traceable.
Organizations must establish an internal governance structure that includes a human-in-the-loop approach to facilitate system accountability. They must conduct privacy and human rights impact assessments as well as algorithmic impact assessments, and a person or persons must be appointed to supervise system development, deployment, and use.
Safe whistleblowing protections must be implemented to shield those reporting non-compliance to an independent oversight body.
The principles provide a framework for risk assessment, system design and development guidance, and the embedding of accountability throughout the AI life cycle. They were crafted to align with international, national, and provincial responsible AI use frameworks like the Ontario Public Service Responsible Use of Artificial Intelligence Directive.
The principles expand on the IPC-OHRC joint statement released in 2023. Patricia Kosseim, Ontario information and privacy commissioner, explained that they set necessary guidelines for organizations to deploy AI responsibly.
“These principles provide a clear roadmap to ensure AI systems are fair, transparent, and accountable. By embedding human rights and privacy protections into the development and use of AI systems, we can build trust and prevent harm,” said Patricia DeGuire, OHRC’s chief commissioner, in a statement.
