38404 (2024). Francisco Hernandez to the Minister of Police

38404 (2024). Francisco Hernandez to the Minister of Police: How, if at all, are the Minister's departments, if any, currently using AI, and what plans, if any, have been developed to manage the risks associated with AI?

Hon Mark Mitchell: I am advised that Police has been using algorithms and AI for many years. Police uses low-end AI features such as image recognition software, speech-to-text, and some automated work flows. Under current policy settings, generative AI tools such as ChatGPT are not approved for any Police use. I am further advised, there is also at least one known example of where Police benefits from third-party/vendor use of AI, which supports the supplier’s in-house processes. This does not interact with any Police data, and does not reside in the Police enterprise IT system. Police has a specific policy for the Trial or Adoption of New Policing Technology, which would cover any future proposals to use AI, this can be found here https://www.police.govt.nz/about-us/publication/trial-or-adoption-new-policing-technology-police-manual-chapter. A Technology Assurance Framework is applied to any such proposal to identify any particular risks and, to develop mitigations. Under this policy and framework, specific uses of technology can be approved by governance groups that include Executive-level sign-off. This process includes establishing appropriate oversight and accountability mechanisms to monitor the ongoing use of those technologies. Police also adheres to the Algorithm Charter for Aotearoa New Zealand, Australia New Zealand Policing Advisory Agency (ANZPAA) AI Principles, and GCDO’s Interim Generative AI guidance for the public service. To support transparency about Police’s use of technologies, including AI, Police publish a periodically-updated technology capabilities List at: https://www.police.govt.nz/about-us/programmes-and-initiatives/police-use-emergent-technologies/technology-capabilities-list. In regard to the Serious Fraud Office (SFO), The SFO’s evidence review platforms include an Artificial Intelligence (AI) machine learning function. Data analysed using AI remains ringfenced within the relevant operating environment. Additionally, the latest version of Adobe Acrobat Pro, which some SFO staff hold licenses for, includes an AI Assistant. The Serious Fraud Office is currently strengthening its internal organisational policy and associated protocols in relation to the use of AI. This policy will be guided by initial advice from the Privacy Commissioner, and the government System Leaders about Public Service use of generative AI tools (Department of Internal Affairs, Statistics NZ, and the National Cyber Security Centre).