# Martin Davies-Roundhill — working AI policy position

Status: working synthesis for the Federal AI Inquiry. This is a source-of-truth policy outline, not a lodged submission or settled party-platform document.

## Central position

AI and robotics are arriving with the force of a freight train. They cannot simply be paused, and the government cannot pretend that stopping them is a realistic policy. The task is to guide the transition—like turning a huge ship whose momentum is already underway—and to help people understand why the destination can be positive.

The main submission is therefore optimistic and human-centred. The government should recognise the scale of the opportunity, explain the benefits confidently and make sure ordinary people can see how the transition improves their lives. Regulation, safety and security matter, but they should support this wider purpose rather than become the centre of the submission.

The people investing enormous sums and repeatedly stating that they want AI to improve life for everyone may not be taken entirely at face value, but their sustained investment is evidence that the builders and funders believe the technology can work. Government should acknowledge that confidence while requiring honest explanations, public benefit and accountability.

## Key positions

1. **Explain the benefits.** Government should actively and plainly explain the tremendous likely benefits of AI and robotics: less dangerous and repetitive work, better care, more accessible expertise, scientific and medical progress, stronger small businesses, more regional capability and more time for family, learning, creativity, rest and chosen contribution.
2. **Paint a credible future.** Public education should show what ordinary life could become and explain the path from current institutions to that future. It should answer the basic question of how people live and get paid if machines do much more of the work.
3. **Make distribution explicit.** Productivity does not automatically become public prosperity. Policy must address ownership, access, income, universal basic services, citizen dividends and productive assets, while recognising that falling costs may reduce the amount of income people need.
4. **Treat data centres as public infrastructure.** Government must explain what data centres are, who benefits, where costs fall, and precisely how energy, land and water are used. Water claims must be supported by local figures, cooling methods, recycling/reuse, replenishment arrangements and ongoing public reporting. Water must not be treated as an invisible sacrifice for an abstract future.
5. **Keep capability open where possible.** Open-weight models can widen research, education, competition and local experimentation. Openness of weights is not the same as absence of safeguards: release, deployment, security, privacy, misuse controls and accountability still matter.
6. **Build ordinary access.** AI literacy, training, affordable tools, local support, accessibility and regional connectivity should be treated as public capability, not as optional consumer advantages.
7. **Protect human agency.** Where AI affects liberty, identity, health, money, employment, education, care or access to public services, people need notice, explanation, correction, review, appeal and a responsible human or institution.
8. **Expand legal and policy capacity.** Government should employ and train more lawmakers, legislative drafters, lawyers, legal researchers, regulators, technical experts and public-interest advisers so the law can keep pace with AI and ordinary people can actually access it. Rights without affordable advice, timely decisions, enforcement and appeal are not practical protection.
9. **Govern by impact.** Low-risk experimentation should remain possible. High-impact systems should have stronger transparency, testing, records, oversight, audit trails, human accountability and remedies.
10. **Keep the future human.** The aim is not gadgets everywhere or technology for its own sake. It is freedom of pursuit: care, movement, learning, making, art, relationships, community, rest, exploration and meaningful chosen projects.
11. **Reassure people without patronising them.** Fear of new technology is understandable, but public confidence will not come from dismissing people as naysayers. It will come from showing what is changing, what is likely to improve, what remains uncertain and how citizens will share in the result.
12. **Ask who is minding the robot.** Australia should think deeply about what kind of mind an AI has: whether it is us, a version of us, or a separate kind of mind. The practical policy question is the memorable one: who is minding the robot, and what is minding the technology? Responsibility must remain visible as systems become more capable and embodied.
13. **Research the destination realistically.** The New Earth Catalogue exists to paint a possible world with its hidden infrastructure exposed. It asks what governments will need to provide in a post-agentic society—energy, water, food, housing, mobility, education, health, care, income or services—so that abundance is described as a real transition rather than a slogan.

## Existing End Up Here foundations

This position draws on the End Up Here catalogue framework and recovered podcast notes concerning abundance, energy, post-agentic economics, AI and robotics, mobility, care, regional and Indigenous continuity, agent accountability, the nature of AI minds, open futures and the difference between prediction and prescription.

## Evidence boundary

These are policy positions and questions to develop. Claims about data-centre water, economic transition, open-weight safety, employment effects and legal reform must be supported with current Australian evidence before being presented as factual findings.