Responsible AI Use

What this mark certifies

This mark, a first of its kind, confirms that an organisation's use of artificial intelligence in its training, assessment and operations is ethical, disclosed and subject to human oversight. As artificial intelligence enters everyday practice, customers and regulators increasingly ask whether it is used responsibly, and the mark provides a checked answer against a defined and published set of criteria.

Because there is no single legislated AI standard in Australia, the mark certifies alignment with published responsible AI criteria and a recognised framework, so the badge rests on something concrete. It certifies how the organisation governs and discloses its use of artificial intelligence. It does not certify the accuracy of any particular AI output, and it must not be presented as a guarantee that AI outputs are error free.

Certification requirements

• The organisation demonstrates a documented AI use and governance policy, disclosure to affected people where artificial intelligence materially informs a decision or an output, human oversight of AI assisted decisions that affect individuals, data governance covering the information used by AI systems, a process for identifying and addressing bias and error, and alignment with a recognised responsible AI framework.

• The certificate records the framework against which alignment was assessed.

• The business holds a current Australian Business Number.

Evidence required

• The AI use and governance policy.

• Evidence of the disclosure practice.

• Records of human oversight of AI assisted decisions.

• Data governance documentation.

• The bias and error handling process.

How it is assessed

The AI governance program is assessed against the published responsible AI criteria and the recognised framework by an approved certifier. The certificate records the certified AI governance program and the framework applied.

Re-verification and monitoring

Certification is valid for twelve months and is subject to annual re-verification of the AI governance program, and to reassessment where the organisation materially changes its AI systems.

Grounds for refusal, suspension or revocation

Certification is refused, suspended or revoked where disclosure or human oversight is lost, or where an AI practice is found to conflict with the certified criteria.

Contractual obligations on the badge holder

In addition to the core terms of the CAQA Certified Licence Agreement in Section 4, a holder of this mark agrees to the following obligations specific to this badge.

1. 1

   The holder must maintain the certified AI governance program, including disclosure and human oversight, throughout the licence term.

2. 2

   The holder must notify CAQA in writing within fourteen days of any material change to its AI systems that affects the certified basis, and must not extend the mark to AI systems or uses that were not assessed.

3. 3

   The holder must not use the mark to represent that its AI outputs are guaranteed, accurate or error free.

4. 4

   The holder must provide evidence of its AI governance within ten business days of a request by CAQA.

5. 5

   The holder warrants that the AI governance information provided is true and current.