Practices that use AI must facilitate processes to obtain and document informed consent and de-identify/anonymise information when using tools that process patient data, according to the sixth edition of the Standards for general practices released by the RACGP on Wednesday.

Identified patient data may only be entered into AI tools if “clinically necessary, explicitly authorised, and supported by documented governance and consent processes,” it states.

Practices must also establish governance processes to ensure accountability and compliance with legislation and document processes that support clinical oversight of AI outputs. Processes are also required to assess and evaluate AI use, including risk mitigation, monitoring, review and quality improvement.

What does all this mean in practice?

The standards offer guidance on how to meet the new criteria, including:

• Patients need to be informed of their right to withdraw consent at any time, and there needs to be a “clear and accessible” way to opt out.
• AI tools need to comply with the Privacy Act and Australian Privacy Principles, as do any third-party providers.
• Practices should develop risk management processes and safeguards to support clinicians.
• Practitioners are accountable for meeting professional obligations regarding AI use.
• AI systems need to be monitored after implementation to ensure they work as planned and to actively mitigate risks.
• Practices should be aware that vulnerable populations may face greater risk of harm from and AI – and should assess and manage these risks.
• They should also be mindful of indigenous data sovereignty.

Read the complete guidance on meeting the criteria here.

The onus is on practices and practitioners

The standards explicitly state that clinicians “are accountable for care decisions supported by AI tools,” later noting that “When AI tools are provided by third parties, practices remain responsible for their safe and appropriate use.”

“Regardless of what technology is used in providing healthcare the practitioner remains responsible for delivering safe and quality care and for ensuring their own practice meets the professional obligations set out in their Code of Conduct, AHPRA says in its guidance. “Practitioners must apply human judgment to any output of AI.”

The standards emphasise the importance of monitoring and reviewing, suggesting strategies including auditing outcomes from AI generated recommendations – such as overdiagnosis, misdiagnosis or missed diagnosis, or tracking feedback or complaints, among others.

College says standards ‘provide flexibility’ for practices

Other new and updated topics include environmental sustainability, defining and planning for your practice, engaging consumers and continuous quality improvement.

RACGP President Dr Michael Wright said shared decision making and patient-centred care were embedded through the standards.

“The standards aim to support practices as they adapt to modern consumer expectations, consider how AI can improve care, implement environmental sustainability, and thrive in an often-challenging financial environment,” he said.

“They’re also strongly outcomes-focused, and don’t prescribe rigid processes. They provide flexibility for practices to implement systems that suit their circumstances, while meeting the expectations of our patients for high-quality care that’s tailored to their needs.”

Time will tell if practice owners and other GPs will agree.

Healthed data from September 2025 found:

• 92% of 312 practice owners surveyed thought the accreditation process needed to be more collaborative and less authoritarian
• 90% thought there were too many unnecessary and burdensome mandatory requirements
• 84% felt the accreditation process lacked flexibility.

The new edition has 31 fewer mandatory criteria than the last edition had. It includes 86 mandatory criteria and three ‘aspirational’ criteria.