Ars Technica: Ontario Audit Reveals AI Scribe Errors in Healthcare

What happened

An audit conducted by Ontario's Information and Privacy Commissioner has found significant inaccuracies in AI-powered medical scribing tools used by healthcare providers. The report, released on May 14, 2026, details instances of fabricated patient information, incorrect treatment recommendations, and even made-up referrals and prescriptions, raising concerns about patient safety and data integrity.

What changed

The audit, which examined the use of AI scribes in Ontario's healthcare system, identified several critical issues. These tools, designed to automate clinical note-taking during patient encounters, are reportedly generating content that deviates substantially from actual patient interactions. Specific examples cited include:

• Fabricated Referrals: AI systems have been observed creating referrals to specialists that were never discussed or recommended by the physician.
• Incorrect Prescriptions: The audit found instances where AI suggested or documented prescriptions that were either inappropriate for the patient's condition or entirely fictitious.
• Misrepresented Patient History: Inaccurate summaries of patient medical history have been generated, potentially leading to diagnostic errors.
• Hallucinated Information: The AI scribes have produced entirely made-up therapeutic interventions or diagnostic codes.

These errors highlight a significant flaw in the reliability of current AI medical scribing technology. The report urges healthcare providers to implement rigorous oversight and verification processes for all AI-generated clinical documentation. The findings underscore the need for more robust validation and error-correction mechanisms within these AI systems before widespread adoption.

Why it matters for agencies

For marketing agencies, this audit serves as a stark reminder of the potential pitfalls of relying on AI without stringent human oversight, particularly in sensitive fields like healthcare. While AI content generation tools can significantly boost efficiency for tasks like drafting ad copy or SEO-focused blog posts, the Ontario audit demonstrates that "hallucinations" and factual inaccuracies are not confined to creative writing. Agencies using AI for client reporting, data analysis, or even generating initial drafts of client-facing materials must implement robust fact-checking and editing protocols. Failure to do so could lead to misinformation, damage client trust, and incur reputational harm, mirroring the safety concerns raised in the medical context.

What to watch next

The Ontario government is expected to release further guidelines for the implementation and oversight of AI in healthcare. Healthcare providers will likely face increased scrutiny regarding their use of these technologies. It remains to be seen how AI scribe developers will address these accuracy concerns and what technical safeguards will be implemented in future updates to prevent similar errors.

---

Source: Your doctor’s AI notetaker may be making things up, Ontario audit finds