Registrar nearly loses trainee position over use of AI

Popular

Read time 2 min

Friday, 18 Sep 2026

Just weeks out from sitting fellowship exams, a GP registrar was facing termination of their registrar position after it was discovered they had used ChatGPT to produce medical documentation. They contacted our Medico-legal Advisory Service seeking support to resolve the issue with their employer, a large corporate practice.

The problem had come to the employer’s attention when the registrar realised they had uploaded a patient referral letter into another patient’s medical record. Wanting to be sure to follow the correct procedure correcting this error, the registrar sought help from the practice manager.

AI solution to a heavy workload

The registrar had decided to try out some of the new AI technology they had heard so many great things about.

Aware that a few of the GPs working for the same practice were already using an AI medical scribe, the registrar started using the AI medical scribe for their consultation notes, but without checking that the practice approved the use of AI tools. The registrar then used ChatGPT to finesse their draft referral letters.

It had simply not occurred to the registrar that, since the draft letters included personal information about a patient, as well as the specialist’s contact details, there were significant privacy issues entering these details into a publicly available externally hosted general-purpose AI tool that operated outside the practice's approved privacy and data security controls.

An investigation by the practice initially concluded the registrar had breached privacy and confidentiality obligations by uploading identifiable patient details into uncontrolled environments.

Practice's policy on AI wasn't clear

Although the management team was considering a trial of the new technology, their current policy stipulated that the use of any AI tools, even those that are specifically designed as medical scribes to generate clinical documentation, was prohibited. This had been noted in the minutes of the meeting when the proposed trial was discussed, and the practice took the view that all staff should have known AI tools were not permitted, as these minutes had recently been circulated.

As part of Avant’s enquiries into the approved processes around documentation, the practice acknowledged their current policy on the use of AI generative tools had not been communicated directly to the registrar, or to other doctors working within the practice.

After this incident, the practice completed their trial and decided that doctors would be permitted to use specific medical scribe tools. This policy is now made clear to all doctors when they join the practice.

Using – and not using – AI in health care

Do not use publicly available general-purpose AI tools, such as ChatGPT, Claude and Copilot, to process patient information. Any clinical information entered into these tools should be completely de-identified, noting that true de-identification can be difficult to achieve and patients may still be re-identifiable from contextual information.

Enterprise AI platforms, such as ChatGPT Enterprise, Claude for Enterprise and Microsoft 365 Copilot, provide additional privacy, security and data governance controls. These platforms may offer greater protection for sensitive information, but their use for patient information should only occur in accordance with your organisation's policies, contractual arrangements, privacy obligations and risk management requirements. Always check your organisation's AI policy before using these platforms.

Check your organisation’s AI policy before using an AI medical scribe and comply with it.

Fixing incorrect records

Thankfully, the registrar had spoken up as soon as they realised their error in entering information about one patient into another patient’s record. This prompt and open approach helped us resolve the dispute with the practice, who accepted a statement of mitigating circumstances.

We were also able to advise the practice on the right way to handle correcting medical records in the situation when Patient A’s information had been incorrectly entered in Patient B’s record:

• Copy the incorrectly entered information from B’s record into A’s record. Add a note in A’s record that this information was incorrectly entered in another patient’s record. Omit B’s details.
• Do not delete any data from B’s record. Instead, annotate this information with ‘Confidential - entered in error’, and flag it so that it will be excluded from referral letters, reports and medical record exports.
• Complete an internal incident report that is kept in the practice’s adverse event log. Include the date the information was entered, the date the error was identified and by whom, patients’ names (or record numbers) and actions taken. Do not copy this into A’s record or B’s record.
• Notify Patient A that some of their health information was incorrectly entered in another patient’s record. Explain the nature of the health information and the steps taken to rectify the error.

Further reading

Fact sheet: Key requirements for medical records

This article was originally published in Connect magazine, Issue 24 and updated on 18 September 2026.