Greens AI plan targets data sovereignty, environmental rules and labelling requirements for AI content

The Green Party has rolled out its artificial intelligence (AI) plan, promising to regulate AI to protect New Zealanders’ creativity, data and jobs and ensure the benefits of AI are shared and not captured just by big technology companies, Green Party co-leader Chlöe Swarbrick said on Monday.

The plan includes sweeping regulation over data sovereignty, environmental regulation and labelling requirements for AI content.

"We don’t work for AI," Swarbrick said. "The Greens will regulate it responsibly, for the public good."

"Nothing about this technology is inevitable, neither the claims of existential destruction by tech billionaires nor the devastation of the financial bubble bursting forecast by economists."

“This is exactly why Government is supposed to exist, to regulate what needs regulating, to ensure benefits are appropriately shared and harms minimised," she said.

The Green Party's AI policy document states that the growing pressures to adopt and embrace AI are unprecedented, but AI is not a force of nature that we have to surrender to.

"The enormous investments being made in AI are cause for concern. Trillions of dollars have been invested in new AI tools, and companies cannot hope to see the returns they expect. This growing bubble presents risks to jobs and economic stability."

Decision making

The document states that "AI must serve people, not replace human and democratic decision-making," and calls for;

• No fully automated decisions in the public service regarding matters that could make a "substantive difference" to people’s lives
• Humans made accountable for AI decisions
• Require public registers of AI systems used by government agencies
• Right to challenge AI-assisted decisions, and to opt out of being subject to AI assisted decision-making
• refuse to purchase or deploy autonomous weapons systems

On data the Greens want;

• Consent standards for AI training data
• Remuneration for intellectual property used for training AI models and fines for breaches
• Specific regulation requiring co-governance and co-development of Māori data and IP
• Transparent datasets used by government AI systems
• Not allowing third party authorisation for public's private information held by the government, or classified government data
• Onshoring public data should be prioritised
• Privacy codes relating to biometric processing and facial recognition technology

On the environment the Greens are calling for mandatory environmental impact assessments for hyperscale AI data centres,  AI facilities having to provide new renewable generation to cover full energy use, water-use limits, and a pause on consenting new hyperscale data centres until a regulatory framework is created.

The party also wants; withholding tax rules applied so tech companies don't offshore their profits,  AI companies to pay local media outlets for data used in their AI summaries, to mandate multiple and interchangeable AI providers when using third-party models, to encourage research towards the development of New Zealand-based and focused large language models (LLMs) to avoid dependence on foreign vendors, and says children shouldn't be forced to use AI.

The policy also calls for labelling requirements for AI-generated content, establish an AI office within the Department of Prime Minister and Cabinet, and algorithm audits for high-risk systems.