Artificial intelligence is rapidly reshaping professional services. Hospitals are using AI-powered “scribes” to reduce doctor paperwork, while law firms and courts are exploring automated transcription systems to process hearings, interviews and testimony more efficiently. The promise is compelling: faster documentation, lower administrative costs and improved productivity. However, new findings suggest that there is considerable risk hidden beneath the efficacy benefits. In high-risk environments such as medicine and law, transcription errors are not simply inconvenient because they can alter diagnoses, affect legal outcomes, and create significant compliance challenges.
A recent audit of 1,000 hours of court proceedings and medical dictations found that AI-generated transcripts contained a “critical error” in roughly 18 percent of the documents analyzed when compared to human-verified records. According to the analysis, these were not routine spelling errors. Errors included incorrect medication doses, altered interpretations of testimony, and omission of critical qualifying words such as “no.” Such errors can fundamentally alter understanding and potentially affect clinical decisions or legal judgments.
The problem of hallucinations
One of the most troubling aspects of generative AI is “hallucination,” those instances where an AI model generates information that appears reliable but is factually incorrect. In transcription systems, hallucinations occur because models try to predict words instead of understanding context the same way humans do.
According to Ben Walker, CEO of Ditto Transcripts, AI systems can perform impressively in controlled environments, but often struggle when faced with real-world conditions. Courtrooms and hospitals rarely offer pristine audio. Multiple speakers may speak simultaneously, regional accents may vary significantly, and technical terminology may be dense and specialized.
The audit identified three recurring problem areas, including overlapping of speech leading to disjointed statements, specialized legal and medical terminology being replaced with common, similar-sounding words, and failure to include negative modifiers such as “no,” creating the opposite meaning of the original statement. These findings align with broader concerns raised by healthcare and legal professionals about the reliability of AI-generated documentation when the consequences of error are severe.
Why health care is particularly vulnerable
Healthcare has emerged as one of the fastest growing applications for AI transcription. Clinicians spend considerable time documenting patient encounters and AI scribes can significantly reduce administrative burden. Canadian studies have suggested that AI documentation tools can dramatically reduce physician documentation and improve workflow efficiency. However, experts have also warned that these systems can miss critical clinical nuances and generate contextual errors when summarizing patient conversations.
This concern is particularly important in Canada, where health care systems face staff shortages and increased administrative pressures. Many organizations see AI as a tool to improve productivity. STILL Canadian privacy experts have pointed out risks associated with AI scribes, including transcription inaccuracies, consent management, data residency concerns, and protection of personal health information.
An incorrectly transcribed prescription, a missing allergy reference, or an altered clinical observation can potentially impact patient care. As a result, privacy commissioners and healthcare regulators in several Canadian provinces have issued guidelines requiring human oversight and review of AI-generated clinical documentation.
Legal systems face similar challenges
The legal sector faces an equally complex problem. Court transcripts serve as official records that support appeals, judicial reviews, and case preparation. An incorrectly transcribed testimony or an omitted statement can change the interpretation of the evidence and create opportunities for legal disputes.
Unlike conventional programs, generative AI systems do not simply convert speech to text. They make inferences about language patterns and fill in information gaps. While this can improve readability, it also presents risks that are unacceptable in environments where accuracy is mandatory.
For this reason, many legal technology experts continue to emphasize verification by trained transcription professionals over fully automated workflows.
The growing use of cloud-based AI transcription services raises important privacy and regulatory questions. Sensitive health data, legal proceedings and confidential customer communications can be processed through external AI systems.
In Canada, health care organizations must navigate federal privacy requirements under Law on Protection of Personal Information and Electronic Documents (PIPEDA) alongside provincial health privacy legislation. Experts note that organizations should carefully evaluate how AI vendors store, process and potentially reuse data. The latest Canadian guidelines regarding AI scribes is heavily focused on managing consent, preventing secondary use of data for AI model training, and ensuring strong governance controls before deployment.
The findings do not suggest that AI transcription should be abandoned. Technology clearly offers benefits in terms of efficiency, supporting workflow and reducing administrative burden. Instead, the evidence points toward a “man in the loop” model.
Canadian privacy experts, health care organizations and legal professionals increasingly view human review as an essential safeguard rather than an optional step. AI can speed up documentation, but humans remain responsible for validating accuracy, context and regulatory compliance.
