Therapeutic Goods Administration: TGA publishes final regulatory framework for software-based medical devices incorporating…

Cresthaven Analytics Intelligence Brief

Headline

TGA publishes final regulatory framework for software-based medical devices incorporating AI and machine learning under the Therapeutic Goods Act 1989

Executive Summary

The Therapeutic Goods Administration has published a final regulatory framework for software-based medical devices (SaMD) that incorporate artificial intelligence and machine learning algorithms, establishing classification criteria, conformity assessment pathways, and post-market change management requirements under the Therapeutic Goods Act 1989 and the Therapeutic Goods (Medical Devices) Regulations 2002. The framework introduces a predetermined change control plan mechanism allowing manufacturers to implement AI model updates within pre-approved boundaries without requiring new conformity assessment for each update.

Key Regulatory Signals

AI-Specific Classification Criteria: The framework establishes classification criteria for AI/ML-enabled SaMD based on the clinical significance of the device's output, the degree of clinical decision-making autonomy, and the transparency of the AI algorithm's decision process, creating a risk-proportionate regulatory approach.
Predetermined Change Control Plan: The PCCP mechanism allows manufacturers to define a scope of anticipated AI model modifications (retraining, algorithm updates, performance optimization) that can be implemented without additional conformity assessment, reducing the regulatory burden of continuous AI model improvement while maintaining safety oversight.
International Harmonization: The framework aligns with the IMDRF SaMD guidance framework and references FDA's AI/ML-based SaMD action plan, facilitating multi-jurisdictional regulatory submissions and reducing the compliance burden for manufacturers seeking simultaneous market access in Australia, the US, and the EU.
Post-Market Surveillance Requirements: AI/ML-enabled SaMD manufacturers must implement real-world performance monitoring programs that track algorithmic performance metrics, detect performance degradation, and trigger corrective action when performance falls below validated thresholds.
Sponsor Obligations: Australian sponsors (manufacturers or authorized representatives) of AI/ML SaMD must maintain technical documentation including algorithm development records, training data provenance documentation, and validation study reports accessible to the TGA throughout the product lifecycle.

Regulatory Delta

The TGA's regulation of SaMD has been governed by the general medical device framework under the Therapeutic Goods Act and the 2021 reforms that brought standalone software within the regulatory scope. The current AI/ML-specific framework represents the first dedicated regulatory instrument for AI-enabled medical devices in Australia and addresses the regulatory gap created by the continuous learning nature of AI/ML algorithms that do not fit within traditional fixed-specification conformity assessment models. The PCCP mechanism draws directly on FDA's Predetermined Change Control Plan framework proposed in 2023 and reflects the TGA's active participation in the IMDRF AI/ML Working Group. Australia's framework positions the TGA alongside the FDA, Health Canada, and the UK MHRA as regulators with dedicated AI/ML SaMD guidance, while the EU's Medical Device Regulation continues to address AI-enabled devices through the general conformity assessment framework without AI-specific provisions.

Materiality Classification

High — TGA final regulatory framework establishing AI-specific classification, conformity assessment, and post-market requirements for software-based medical devices in Australia, with direct implications for manufacturers seeking Australian market access for AI/ML-enabled health technologies.

Time Horizon

6 months — Framework effective 6 months from publication; manufacturers of AI/ML SaMD must complete conformity assessment under the new framework prior to ARTG inclusion or renewal.

Intelligence Outlook

Monitor the TGA for implementing guidance on PCCP preparation and submission requirements. Track the IMDRF AI/ML Working Group for international harmonization developments. Assess interaction with the Australian Government's AI Ethics Framework and any broader AI regulation legislation.