Our Principals Based AI Policy

These are the rules that govern our use of Artificial Intelligence.

At AI New Zealand, we will use Generative Al in accordance with these principles:

1. We won't incorporate confidential, sensitive or personal information into a publicly available Generative Al tool.

(a) We will use publicly available Generative Al tools as long as our use is consistent with this policy and appropriate in the circumstances.

(b) We won't enter any personal, sensitive or confidential information into a publicly available Generative Al tool unless we have the relevant person or entity's express and informed consent.

(c) We will make sure the settings of any Generative Al tools we use are managed to safeguard our information (including clients' information) to the extent appropriate. This may include (where appropriate) disabling training from our data. It will always be appropriate to disable training on confidential, sensitive or personal information. If information is anonymous, de-identified or publicly available, we will consider any risks before permitting training, and will permit training unless any risks are identified.

(d) When we procure or develop any enterprise/private Generative Al tools, we will do appropriate due diligence (including security and privacy impact assessments where relevant) to ensure they are secure, legally compliant and ethically sound.

(e) We will make sure our people are trained on the use of Generative Al tools, so that we use them in a secure, legally compliant and ethical way.

2. We will check the outputs of Generative Al before using them.

(a) We will also ensure the outputs are not the only source we use for any work product.

(b) If using a substantial part of the output of Generative Al in our work, we will:

• check the terms of the Generative Al tool to make sure this is permitted;

• check any knowledge cut-off dates for the tool;

• check the content itself to ensure it has not been copied from another source; and

• ensure the use of Generative Al is appropriately disclosed/referenced in our work.

3. We will be transparent about our use of Generative Al - with each other and with clients and stakeholders.

(a) We will always let the senior person on a matter know if, and how, we want to use Generative Al on that matter. We will do this in advance of any use so we can discuss the implications and agree an approach consistent with this policy.

(b) We will proactively disclose the use of Generative Al tools to our clients, in advance of any use, where we believe there is additional risk involved in the particular use of Generative Al.

4. We will communicate in our own voice.

(a) Including, as much as possible, where our voice is portrayed through a digital avatar.

(b) We can use Generative Al as a starting point for developing ideas and arguments, but won't solely rely on it to produce detailed content or advice.