Government AI policy already 'overdue,' Opposition Leader warns
Opposition Leader Tina Browne says artificial intelligence is already in use across the Cook Islands. Her message is blunt: without clear rules, jobs, rights, and trust in public institutions are at risk.
That's not a theoretical worry. Agencies are trialling tools for drafting, analytics, and service delivery. The country needs a policy that sets boundaries, protects people, and speeds up safe adoption-now, not later.
Why it matters for government teams
- Job security: automation without a reskilling plan leads to quiet cuts and low morale.
- Fairness: biased data can skew decisions on benefits, licensing, and hiring.
- Privacy: staff may paste sensitive data into public chatbots without guardrails.
- Trust: opaque systems erode confidence if people can't see how decisions are made.
What a practical AI policy should include
- Scope and risk tiers: define what counts as AI, and set stricter rules for high-impact uses (e.g., eligibility, policing, health).
- Data protection: clear rules for personal data, consent, retention, and approved datasets.
- Human oversight: a named decision-maker for high-stakes outcomes, with the right to review and reverse.
- Algorithmic transparency: plain-language notices to the public and records of model versions, prompts, and outputs.
- Impact assessments: require bias and safety checks before deployment, with consultation from staff and communities.
- Procurement standards: vendor obligations for risk testing, security, uptime, audit rights, and incident reporting.
- Security and resilience: controls for model misuse, prompt injection, data leaks, and third-party risks.
- Redress and complaints: simple ways for people to challenge AI-assisted decisions and get timely responses.
- Governance: a small central AI function, named leads in each ministry, and a public register of AI systems.
- Records and audit: keep logs of AI-assisted decisions to meet public records laws and support audits.
- Skills and change: budgeted training, job redesign, and a plan to measure productivity and service quality.
Immediate steps agencies can take
- Inventory: list every AI tool in use or on trial, including shadow IT. Note purpose, data used, and owner.
- Appoint an AI lead: give them authority to pause risky uses and approve new ones.
- Set usage rules: no confidential or personal data in public chatbots; approved models and accounts only.
- Do a quick risk check: test for bias, factual errors, security gaps, and prompt injection on critical workflows.
- Tighten procurement: add AI clauses-data ownership, model change notices, security attestations, and exit terms.
- Train staff: baseline training on safe prompting, privacy, and citation. Start with frontline and policy teams.
- Communicate with the public: explain where AI is used and how people can opt out or appeal.
Jobs and capability
AI will shift work, not just cut it. Treat this as workforce planning, not a gadget rollout.
- Map tasks, not roles: identify repetitive steps that can be assisted while keeping human judgment.
- Redesign roles: move saved time into inspections, case reviews, and community engagement.
- Reskill: build prompt fluency, data literacy, and model oversight skills across teams.
- Partner early: involve unions and staff reps in policy, pilots, and evaluation.
Standards you can lean on
Don't start from zero. Borrow what's already proven and adapt it to local needs.
- OECD AI Principles for high-level guardrails.
- NIST AI Risk Management Framework for practical risk controls.
What this means for the Cook Islands public sector
The country's size is an advantage. A lean policy, a short list of approved tools, and a clear register can be set up in weeks, then improved with feedback.
Delay carries real costs-quiet misuse of data, untested models in sensitive areas, and public pushback when things go wrong. Setting the rules now protects people and gives agencies confidence to use AI well.
Get your team AI-ready
If you need structured upskilling for government roles, see curated options by role and skill level here: Courses by Job. For recognised pathways, explore popular AI certifications.
Tina Browne has raised the flag. The next move is simple: publish interim rules, launch a public register, and start training. From there, iterate with evidence.
Your membership also unlocks:
