Ontario’s use of artificial intelligence is facing new scrutiny after the province’s auditor general found weaknesses in government oversight, security controls, and testing standards for AI systems. According to reporting by CP24 and findings from Ontario Auditor General Shelley Spence, the audit reviewed AI use across the Ontario government from January to November 2025 and found that the province’s strategy was still missing several important safeguards.
One of the key concerns involved AI note taking tools used by some family doctors and health care professionals. The auditor found that some tested AI scribe systems produced serious mistakes, including recording the wrong medication, adding information that was not mentioned in simulated conversations, and suggesting treatment steps such as therapy referrals or blood tests when those actions had not been discussed. The findings raise concerns about how much trust should be placed in AI generated medical notes without careful human review.
The report also found that accuracy was not given enough weight when the government assessed approved vendors for AI scribe tools. The auditor warned that this could lead to the selection of systems that create inaccurate or biased medical records or fail to properly protect sensitive health information. The report recommended that users be required to confirm they have reviewed AI generated notes after appointments.
Ontario Public and Business Service Delivery and Procurement Minister Stephen Crawford said doctors still make final decisions on prescriptions and treatment. He also noted that the issues flagged by the auditor were found during testing rather than confirmed real world use. Crawford said AI note taking can save doctors an average of about five hours per week in clerical work, while emphasizing that the goal is to support workers rather than replace them.
The audit also raised concerns about government staff accessing unsafe AI websites from work devices. According to the report, between April and August 2025, about 12,000 Ontario Public Service staff accessed roughly 400 AI related websites. The auditor found that about 60 percent of those sites were considered unsafe or unsecured, and some were not related to work. The report warned that staff could unintentionally upload personal information, health card numbers, driver’s licence details, credit card information, contracts, invoices, or other sensitive data to unapproved platforms.
The auditor also noted that only three percent of Ontario Public Service staff had completed the government’s responsible AI training course as of August, even though the course had been available since January 2024. Microsoft Copilot Chat was identified as the only approved secure generative AI website for staff, but it represented only a small share of AI site usage. The government says it has since taken steps to block unsafe AI websites on public service devices.
Another area of concern involved Ontario’s planned Document Verification Service, which is expected to help people confirm their identity online when accessing government services. The system uses live video and facial movements to verify that a person is real. However, the auditor found that testing used a sample size that was too small and not representative of Ontario’s diverse population, creating the risk that some demographic groups could face higher rejection rates or delays.
The auditor said the province accepted most of the recommendations aimed at improving AI security, oversight, fairness, and reliability. As AI becomes more common across health care and public services, the findings show that Ontario must balance innovation with strong safeguards to protect privacy, accuracy, and public trust. More Canada news and public policy coverage can be found at https://www.weeklyvoice.com and https://www.weeklyvoice.com/category/canada.
