In July this year, the Australian Government announced the creation of an Office of AI within the Department of Prime Minister and Cabinet. The announcement matters not simply because of the technology, but because it recognises that artificial intelligence requires a national conversation about governance, standards, and public trust.

For the past several years, the conversation about AI has largely focused on capability. We have marvelled at its ability to generate content, respond to questions, analyse data, and automate tasks. Yet as AI becomes more powerful, a different set of questions are urgently needed.

Artificial Intelligence Needs Governance

Artificial intelligence is often discussed as a technical capability. In reality, it is mostly an information challenge - one concerned with how knowledge is created, organised, interpreted, and trusted.

AI influences what information is surfaced, how knowledge is represented, what perspectives are prioritised and, increasingly, how people will understand the world around them. The challenge facing society is therefore not simply creating more information. It is ensuring information is used responsibly, with clear answers to questions about trust, bias, transparency, equity, evidence, and public confidence.

These are not technical questions. They are questions of governance, stewardship, and accountability.

A Missing Perspective?

As governments establish AI offices, standards and regulatory frameworks, the conversation will likely involve the usual suspects - technologists, policymakers, lawyers, and consultants.

Yet AI is fundamentally an information service, and that means the governance conversation needs to extend beyond technology alone.

Questions of information credibility, evidence, access, transparency, provenance, and trust have been studied for generations by researchers, educators, archivists, records managers and librarians. Long before debates about algorithmic bias, information professionals were grappling with classification bias. Long before concerns about AI hallucinations, they were teaching people how to find information, critically evaluate sources and challenge assumptions.

These are not new challenges. They are longstanding questions about how knowledge is organised, evaluated, and stewarded, now being amplified through systems that operate at unprecedented speed and scale.

Sovereign AI or Trusted AI?

Australia is increasingly discussing sovereign capability in AI. Much of that discussion focuses on infrastructure, platforms, data, commercial innovation, and technical expertise.

These investments are important, but sovereign capability is not only about owning technology. It is also about shaping the values that underpin it.

For the past few decades, much of the world's information infrastructure has been built by private technology companies. Search engines have been optimised for popularity. Social media platforms have been optimised for engagement. The results have been extraordinary (as have the profits), transforming how we communicate and discover information while also exposing deep challenges around misinformation, bias, trust, ethics, and public accountability.

As Australia invests in AI, there is an opportunity to ask a broader question: what if we invested as heavily in trusted AI as we do in AI capability?

What if sovereign capability also meant investing in the education, research and information sectors that help societies create, evaluate, and trust knowledge? What if expertise in information stewardship was considered just as important as expertise in technology?

We cannot know whether the digital world would look different had those investments been made earlier. But it is reasonable to ask whether systems designed around evidence, understanding and public value might have produced different outcomes than systems primarily optimised for attention, engagement, and commercial growth.

Building Trusted AI

Imagine information systems designed to maximise understanding rather than engagement. Systems that make evidence visible, distinguish between established knowledge, emerging understanding, and speculation. Systems that expose uncertainty rather than conceal it and help people develop judgement rather than simply consume answers.

Most importantly, they would be systems that maintain a clear line of human accountability. AI can generate recommendations, summaries, and confident conclusions, but it cannot be accountable for the consequences of those decisions. Only people can.

One of the defining governance challenges of the AI era will be ensuring that responsibility does not disappear into algorithms, vendor hype, or black-box systems. There must always be a clearly identifiable human being accountable for how this technology is applied, how decisions are made and how harms are addressed.

Transparency, evidence, and stewardship all matter. But accountability is what ultimately converts trust into action.

Without accountability, intelligence - whether artificial or human - cannot be governed.