Alfred Health

Patients, health consumers from all over Australia- people, children, teenagers and parents ask me "what can I do" to have say in how AI affects me? With metrics of trust in insitutions falling in many countries, ordinary everyday people are retreating from participating in public conversations, some are not returning to hospitals where there has been a breach of their data privacy. The disempowerment is spreading as they read about another automation or update or technology that they dont choose, is foisted on them. The dialogue will be a success when it provides the 'consumers' a list of things thet, individual citizens can do and expect . The pace of change takes ordinary people out of control and this accelerates divisions already sowing across the globe.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Safe, secure and trustworthy AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

AI in healthcare risks amplifying existing vulnerability unless governance is explicitly designed to protect those with the least power in the system. The cross cutting themes need to ensure the AI is used to serve consumer health needs (and theirefore reformand transforamtion) and not disempower them. Trust is the infrastructure of health systems and it is under pressure. Public trust in healthcare has been built over centuries from integrity, competence, and accountability. It is not only a cultural asset; it is the mechanism by which health systems function. As a board director of public health and non-profit organisations, my research consistently shows that without trust, data slows or stops, compliance declines, and the populations most in need disengage entirely. As AI accelerates through health systems, this infrastructure is at risk of being bypassed rather than strengthened. Studies show nearly half the public actively opposes automated decision-making in healthcare, not because they have considered and rejected it, but because they don't understand how it works or believe it will serve their interests. This is a governance failure, not a communication problem. The remedy is inclusion: people must be part of the design, build, and use of AI systems that will make decisions about their care. Australia's own TGA review confirms that current medical device regulations are broadly suitable for AI, but that targeted reforms are urgently needed. The most critical vulnerabilities lie in adaptive and continuously-learning AI, post-market surveillance, transparency obligations, and definitional ambiguity around regulatory responsibility. The pace of AI development in medical research continues to outstrip regulatory response. This gap falls hardest on those least able to advocate for themselves. Board directors and audit committees in health institutions need tools to treat AI governance as a fiduciary responsibility not a delegated technology question as they do with protocols for clinical trials. Accountability for how AI systems affect patients, especially vulnerable populations, belongs at the highest levels of institutional leadership. The question before this Dialogue is who does AI work for. In Australia, 560,000 people remain without internet access; often those with the greatest health need and the least power to contest a system that fails them. We must make these technologies powerful for everyone or they will deepen the very inequities health systems exist to address.

The impacts of AI governance decision making on children, our future adults and on seldom heard groups of people, those on the margins of our societies have no real choice or redress like an AI Ombudsman, an automated decision making parliament for children, because the decisions made in boardrooms and political corridors don't include them. A child doesnt know how to exercise the power under the UN Rights of the Child or the HUDERIA framework. They are flat out negotiating with parents and teachers and peers about hour by hour rules of life. The biases we fear being entrenched by AI models also apply to poor people, those without a phone or computer or access to information about what is coming their way. Nothing about us without us was the catch cry of the movement for disability rights, but it originates in the struggle against the divine right of Kings in Eastern Europe. I look to the UN and the EU to wage this battle again, for every powerless person I do my best to work for, everyday.

As a director on two large national public health boards in Australia I am concerned Australia's approach to AI and automated decision-making in public healthcare is sounding like strong high-level principles but a lacking clear, operational guidance for hospital boards and directors. Governance is fragmented across privacy law, medical device regulation and voluntary frameworks, leaving uncertainty about accountability, oversight and risk management when AI influences clinical or administrative decisions. In practice, this means many health services are deploying AI without consistent requirements for registers, pre-deployment assessment, monitoring, or patient disclosure. Boards carry fiduciary and clinical governance responsibilities for these technologies, yet often lack the capability, standards and reporting frameworks needed to exercise effective oversight. The consequence for patients is a widening gap between expectation and reality: AI is already shaping care, but transparency, consent, equity safeguards and avenues for redress remain underdeveloped. For children, for people with disability, for those who do not have the language, confidence or power to question the system, this gap is not abstract. They cannot interrogate an algorithm, challenge a decision pathway, or meaningfully consent to the secondary use of their data. They rely entirely on the integrity of the system to protect them. If governance remains unclear and accountability diffuse, we are asking the least powerful people in our community to bear the greatest risk of technologies they neither chose nor control. Without clearer direction and enforceable governance models, there is a material risk that AI adoption will erode trust, obscure accountability and expose patientsparticularly those most vulnerable to avoidable harm at scale.

Provide an internationally useful tool set for safe and ethical decision making processes and workflows for all governance leads- Chair, board, executives, MP's, clinicians on. A set of questions and stages that help make decision making and role accountability very clear. Public good and public services boards that are not mandated inside government, non-profit organisations, services, hospitals for example have fiduciary duties to ensure these technologies are intrduced and monitored responsibly. Few know how. It requires strategic oversight and the courage to withstand vested pressures even from political spheres (which I know about personally as the spouse of a former federal political leader in Australia) to create the conditions for safe and responsible citizen involved healthcare reform and digital transformation for people who need it most.

Most patients can see what AI might offer. Better diagnosis. Fewer delays. A system that doesn't lose things. When I ask people, they will often tell me they are open to it especially if it means faster answers, greater accuracy, or care that reaches them where they actually live rather than where the system finds it convenient to deliver. That openness is a privilege for me to hea But we arent worry. People are afraid of losing the human connection at the centre of healthcare that relationship with a clinician who knows you, not just your record. They are worried about safety. About what happens to their data. About whether an AI trained on populations that didn't include people like them will serve them well, or whether it will quietly encode the same inequities that already exist in the system and call it progress. And here is what the evidence tells us clearly: trust is not given. It is earned through specific things. Whether the system is accurate and reliable. Whether it can explain itself — not in technical language, but in terms a real person can evaluate. Whether there is a human being in the loop, accountable, present, who can be questioned and who can be wrong and who knows it. These are not unreasonable demands. They are the same things we have always asked of good medicine. Younger patients tend to be more comfortable with AI. More educated patients, more digitally experienced patients. Those encountering it in familiar, lower-stakes settings — a radiology result, a screening tool — tend to accept it more readily than those meeting it in high-stakes, emotionally charged moments. None of that should surprise us. What it tells us is that where we introduce AI, and how, and to whom, matters as much as whether we introduce it at all. The leading guidance from the WHO, the EU AI Act, HUDERIA, NIST, the FDA, the NHS converges on the same principles, and they are not complicated. Protect patient autonomy. Be transparent. Be explainable. Keep humans meaningfully in charge. Make fairness a design requirement, not a hope. These are principles that every decent health system already claims to hold. AI asks us to mean them technically, not just rhetorically. The implementation evidence is equally clear. Co-design works. When patients and communities are involved in how AI tools are built and deployed, not just consulted after the fact, acceptance improves and outcomes are better. Consent processes that are genuine and that actually explain what is happening and why, in language that makes sense to the person, build confidence rather than eroding it. Ongoing audit, feedback loops, and the willingness to say something isn't working and change it: these are the marks of a system that takes its obligations seriously. We are also honest about what we don't yet know well enough. In culturally and linguistically diverse communities, we have far less evidence, far fewer co-designed tools, and far too many assumptions that what works in one population will translate to another. That gap is not a technical oversight. It is a justice issue. If AI in healthcare improves outcomes for some patients while leaving others further behind, we have not advanced medicine. We have automated inequity. The research agenda from here is clear. We need longitudinal studies — not snapshots. Multi-centre work, across diverse populations, measuring not just clinical accuracy but trust, understanding, and health equity as outcomes in their own right. We need to know whether AI is actually changing what matters to patients, not just what is measurable for researchers. And we need to take the legal and ethical questions with the same seriousness we take the clinical ones. Data privacy. Algorithmic bias. The digital literacy gap that means some patients will be able to navigate AI-mediated care confidently while others are left unable to advocate for themselves. Inclusive design is not optional. Neither is regulation that keeps pace with the technology.

citizens juries would be valuable. Too often NGOs dilute voices unwittingly. Vulnerable people need to be called to participate. a caravan of ideas to places and people hard to access An accountability repository a clinician led opt out mechanism for vulnerable people

All the consumer health group members, children, the elderley, people in the bush, Aboriginal Australians, those who dont attend health check ups, those without access to technology.

real time polling and sentiment anlaysis from participants published as they engage not edited or produced after the fact. This is a powerful model

UTS Human Technology Institute, Sydney IEEE Ethically Aligned Deisgn Sandeep Reddy work on Governance Framework for Global Digital Health Transformatiuon AI Verfiy SIngapore