Privacy Commissioner warns companies using generative AI tools that they are bound by privacy laws and checks procedures for organizations or businesses planning to use AI tools in combination with personal information I made a list.
An official guideline, this checklist includes human review before use to ensure that AI output is accurate, and to ensure that personal information is not inadvertently disclosed by AI. was
Secretary Michael Webster also said that all government agencies using systems that might capture personal information of New Zealanders to create new content should be made aware of the implications of the use of AI before launching. He said he expects to consider it.
AI’s use of personal information of New Zealanders is regulated under the Privacy Act 2020.
The move comes after revelations that the Kuomintang has already started using generative AI tools to create ads on social media.
The images included AI-generated images of a hooded man committing a robbery, an anxious homeowner, and Pacifica medical workers.
Webster outlined seven pieces of advice to help companies and organizations considering using generative AI.
1. Get Senior Leadership Approval
He said companies and organizations should involve senior leaders and privacy officers in deciding whether or how to deploy generative AI systems.
2. Consider whether generative AI is necessary and appropriate
3. Conduct a privacy impact assessment
This included seeking feedback from affected communities and groups including Maori.
4. Be transparent
This includes communicating in plain language to customers and clients that the tools are being used and how privacy risks are being managed.
Webster said this is essential to maintaining trust and upholding the social license of organizations to use AI.
“Generative AI is a new technology and many people will be uncomfortable using it or not understand the risks to them,” Webster said.
5. Create procedures for accuracy and personal access
This includes creating procedures to ensure that information generated by AI is accurate and to respond to requests from individuals to access and correct their personal information.
Robert Kichin/staff
Privacy Commissioner Michael Webster said his team would investigate violations of privacy laws through the use of AI.
6. Ensure Human Review Before Acting
This includes having a human review the AI output before the information is acted upon.
7. Prevent personal or confidential information from being retained or disclosed
“Do not enter personal or confidential information into a generative AI tool unless the information entered is explicitly confirmed not to be held or disclosed by the provider,” said Webster.
“An alternative approach is to remove any information from the input data that allows re-identification. We strongly caution against using sensitive or sensitive data for training purposes.”
He said his office will investigate whether the parties feel they are violating the Privacy Act 2020 in their use of generative AI tools.
He previously sent a letter to government agencies outlining their warnings against the premature use of generative AI without proper evaluation, calling for a government-wide response to the growing challenges posed by the tool. hinted at the need to respond.
