LCO Submission on Trustworthy AI -- Final -- June 4 2021

In light of the above, the LCO recommends that the provincial government commit to not procure, operate or deploy AI systems in the areas set out in Annex III prior to the development of a comprehensive, legislated Trustworthy AI Framework.5 Recommendation #2: The provincial government Trustworthy AI Framework will be established in legislation and regulations. [...] The appropriate elements of the LCO’s recommended legislative provisions are set out in the LCO’s Regulating AI paper.6 Arguably the most significant risk associated with government use of AI and ADM systems is their potential to amplify existing bias and discrimination.7 As a result, the LCO recommends that Ontario’s Trustworthy AI legislative and regulatory framework address how to effectively p. [...] The LCO’s The Rise and Fall of Algorithms in the American Justice System: Lessons for Canada report discusses the risks of AI and automated decision-making systems in the criminal justice system at length. [...] In the LCO’s view, the experience of the New York City Automated Decision-making Task Force demonstrates why public education is necessary for successful public consultations.12 The LCO believes it is important the public be invited to provide input before, during and after the initial development and regulation of AI and related systems. [...] The LCO believes the provincial government should prioritize the identification of prohibited and high- risk AI systems and the development of an AI impact assessment tool.