AI use cases in the OPS

URL: https://data.ontario.ca/dataset/29c77a6b-22d3-434e-94d5-51b25d8d8cc0/resource/b8778f78-f1dc-4f95-9bd1-2420bb4cbc11/download/4._cas_dutilisation_de_lia_dans_la_fpo_1.csv

The Government of Ontario is committed to the safe and responsible use of Artificial Intelligence (AI) to improve programs and services across government, as outlined in Ontario’s Trustworthy Artificial Intelligence (AI) Framework.

Under this framework, the Responsible Use of Artificial Intelligence Directive sets out the requirements for the transparent, responsible and accountable use of AI. One of these requirements is to publish a list of AI use cases. This dataset includes AI use cases implemented within the Ontario Public Service (OPS).

The dataset will be updated periodically as additional use cases become available for publishing. The information provided on the AI use cases includes the following:

• AI use case name
• Ministry
• Use case description
• Launch date
• AI capabilities
• Autonomous
• Human in the loop
• User base
• What data is used?

Artificial intelligence has been around for many decades, and its definition has evolved over time. Capabilities which were considered AI in the past may not be considered AI today. The dataset excludes:

• Non-OPS AI use cases: AI use cases implemented in Agencies, Boards, and Commissions (ABCs) and the Broader Public Sector (BPS).
• Sensitive AI use cases/FIPPA exemptions: Use cases where the disclosure would reveal information that should not be disclosed for legal, security, law enforcement, confidentiality, privacy or commercial sensitivity reasons, as identified and documented by expert assessors (e.g., legal counsel, cyber security) and in alignment with existing exemptions in the Freedom of Information and Protection of Privacy Act (FIPPA).
• Emergency use: AI systems deployed during emergencies (e.g., as defined in the Emergency Management and Civil Protection Act). These AI use case should be posted after the emergency has ended.
• Legal restrictions: Use cases that cannot be disclosed by law.
• AI systems which are not adaptive or autonomous: Systems that follow set rules without learning or making complex decisions, resulting in predictable outcomes.
• Planning/design stages: AI systems that have not yet been deployed into production.