Overview
NAIC launched the Voluntary AI Safety Standard (VAISS) in September 2024.
In 2025, we’re looking to extend the standard to include additional practices and guidance for AI system developers. These will be aligned with the 10 guardrails laid out in the first version (V1).
Version 2 (V2) of the VAISS will also provide guidance on labelling and watermarking of AI content and incorporate feedback from industry on V1 to improve practices and guidance for deployers.
We will be holding online consultations in late January and early February 2025. We encourage individuals and organisations to express their interest in taking part.
Version 2 scope
Our 2025 consultations will include discussions about:
Labelling and watermarking: requirements for developers and deployers to implement labelling and watermarking, communication to end-users, technical options available and alignment to international standards.
Enhancements to 10 Guardrails: detailed requirements for AI model and system developers, focusing on developer best practice aligned to the voluntary guardrails.
Procurement guidance: updates to provide more details, supported by a standalone procurement guide.
VAISS V2 consultations
There will be a maximum of 9 virtual consultation sessions. We are planning for these sessions to take place between 27 January and 14 February 2025.
Sessions will be grouped to ensure a robust discussion. Examples of groupings might comprise (but are not limited to):
Australian AI start-ups
Australian AI Tech companies
AI-enabled product and service providers
government agency
civil society
not-for-profits (NFPs)
academia and research organisations.