Data and modelling
This guidance outlines responsible artificial intelligence practices for businesses regarding data collection, training, and modelling. It covers considerations for fit-for-purpose data, legal and ethical data use, intellectual property, indigenous Māori data sovereignty, and model efficacy and testing. Practical tips and case examples are provided to help businesses mitigate risks such as bias, privacy breaches, and copyright infringement.
Summary and content fields are drawn from the document text and are pending editorial review.
Who issued it
Who it governs
This is for businesses, developers, and deployers seeking guidance on responsibly managing data and modelling practices for artificial intelligence systems.
What the document contains
The document covers data and modelling considerations for responsible AI, including training data quality, bias mitigation, privacy impact assessments, intellectual property rights, indigenous data sovereignty, and model testing and monitoring.
It does not contain legally binding enforcement mechanisms, statutory penalties, or specific technical source code definitions.
- Prioritise data quality from the start to ensure reliable, unbiased, and accurate AI outputs.
- Businesses must evaluate privacy, confidentiality, and legal risks when collecting and processing training data or user inputs.
- Copyright permissions and fair licensing are critical when sourcing training data, especially for generative AI systems.
- Māori data requires specific consideration, consultation, and safeguards to respect indigenous data sovereignty and cultural protocols.
- AI models should be actively monitored, tested, and documented throughout their lifecycle to prevent data drift and performance degradation.
· Fit-for-purpose data · Scenario B: BigBuild · Legal and ethical data · Sensitive, proprietary, or personal information · Ownership and intellectual property rights · Options to ethically source datasets including copyright works · Māori and other indigenous data · Model efficacy
How it scores
Each score is backed by a quote verified against the document text. Editorial review has not yet confirmed them.
Accountability
3/5Clear responsibility for AI outcomes
Document (or require documentation on) how the data was sourced or collected, any modifications made and known biases or accuracy levels.
Tips
screening score · 2026-10-24
Fairness
4/5Non-discrimination and equitable treatment
Data fields containing sensitive or protected attributes should be defined and handled appropriately to minimise harm.
Tips
screening score · 2026-10-24
Human Dignity
3/5Respect for the inherent worth of persons
Improper data collection and processing can mean potential privacy and confidentiality breaches, intellectual property rights violations, or potential data sovereignty and human rights impacts.
Fit-for-purpose data
screening score · 2026-10-24
Human Oversight
3/5Humans remain in control of consequential decisions
Following a risk assessment, BigBuild puts in place some risk mitigations, including being transparent with applicants about AI involvement in the shortlisting process, and implementing a human reviewer for all shortlisted applications.
Scenario B: BigBuild
screening score · 2026-10-24
Privacy
4/5Protection of personal data and private life
Practices such as data anonymisation, encryption and secure storage can be employed to help protect personal, confidential and other sensitive information.
Sensitive, proprietary, or personal information
screening score · 2026-10-24
Transparency
4/5Openness about how AI systems work and are used
Developers can produce an AI model (or service) card detailing model purpose, data sources and provenance, training methodologies, performance metrics
Tips
screening score · 2026-10-24
Enforcement
1/5Consequences and compliance mechanisms with real force
Consequences may be reputational, financial, punitive, or obstructive (such as via compliance, cease and desist, or asset freezing orders).
Fit-for-purpose data
screening score · 2026-10-24
Impact Assessment
4/5Required assessment of an AI system's risks/impacts before or during use
Conduct a Privacy Impact Assessment at the design stage to support a privacy-by-design approach.
Sensitive, proprietary, or personal information
screening score · 2026-10-24