The Humanware Institute analysed the evidence of all 575 contributors to the Australian Parliament's AI inquiry. It finds support for AI is broad and the argument is about the terms.
5 November 2026 - Sydney, Australia: AI agents have begun making payments on Australian cards. But who carries the loss when a consumer's own agent acts within the permission it was given but beyond what the person intended?
Nine financial-system institutions gave evidence to Parliament's the Australian Parliament's AI inquiry: three major banks, their peak body, four regulators and Treasury. Asked who answers when an agent acts for a consumer, the banks pointed to the companies that build agents, the consumer regulator pointed to Treasury, the markets regulator pointed to the law as it stands, and the Productivity Commission pointed to a review not yet done. In the evidence examined, none said who carries the in that case. Existing law applies to AI; how it applies to that case remains contested or untested. That is one finding of AI on Australia's Terms: What Parliament's AI Inquiry Reveals About Power, Value and Accountability, a 106-page report published today by the Sydney-based The Humanware Institute, with Kate Cooper as principal author. AI did the first reading and coding of the evidence, under a written protocol and human checks set out below. The report analyses the public record of the federal Joint Select Committee on Artificial Intelligence: all 564 public submissions, spanning 6,158 pages; the ten hearing days held between 18 September and 22 October, totalling 53.4 hours; and the 66 answers to questions on notice published by 28 October. Together these represent 575 distinct contributors and separately, the report maps 158 Australian AI initiatives, partnerships and rules, and compares the responses of eight other governments. Its central finding is that contributors broadly support AI, and most of them set terms: skills to use it well, a say before it changes their work, a fair hearing when it decides about them, control over what is built from their work and voices, and someone who answers when it acts on their behalf. The evidence reviewed identifies no one who owns delivery of those terms as a whole.
"The people and organisations who wrote to Parliament mostly said yes to AI. Then they set terms," said Kate Cooper, Founder and Chief AI Officer of the Humanware Institute. "Skills, a say, a fair hearing, and someone who answers when AI acts for them. The committee's job on 30 November is to turn those terms into conditions people can rely on."
Key Findings:
- Of 575 contributors, 367 support AI. 290 of them attached conditions. Organisations were more supportive than individuals: 109 of 207 individual contributors were cautious or opposed.
- AI skills and literacy was the most common request, raised by 138 contributors, about one in four, and 35 more than the next most common request. MEGT, which works with about 44,000 employers, told the committee that only 3.7 per cent of employers providing AI training use nationally recognised training.
- 82 contributors asked for notice, reasons and a human review that can change an automated decision. 61 asked for clear liability across the AI value chain. Mastercard reported Australia's first authenticated agent transactions in January 2026, in trials on two major banks' cards. The Reserve Bank's Review of Payments System Regulation, which covers agentic commerce, announces conclusions in 2028.
- 17 of 18 unions asked for consultation before AI changes work, compared with one of 35 banks and other businesses. These are requests made to the inquiry, not employer practice. The Finance Sector Union told the committee only two employers in recent years have been prepared to consider an AI consultation clause.
- A voice actor told the committee in his submission that he found his cloned voice earning money on a major video platform, sent the platform more than 100 emails over 12 months, and approached seven agencies and services without finding a way to enforce his rights. A voice is not a copyright work, the report notes, so copyright offers no remedy. The ABC told the committee it has not been paid by any frontier AI developer for training use.
- Of the 195 submissions that addressed the question, about 190 said Australian hosting alone does not guarantee public benefit. UNSW told the committee, citing a register of customers for planned AI compute, that future capacity is already fully reserved and most tenants are offshore companies.
- Australia's largest AI commitments have waited two to three years. 17 of the 20 major model releases on the report's timeline came after the government committed to a law on automated decisions in November 2023. That law has no date.
The terms people asked for Skills, then work design. Access alone does not deliver value, contributors said. Westpac told the committee the greatest constraint on adoption is no longer access to technology but workforce capability, and The AI Accelerator described cutting the cost of checking an insurance sales call from about $75 to about $3 by redesigning the work around a system it built. Human review with authority. Aged-care assessors could see an automated funding decision getting their clients wrong and were barred from overriding it, according to evidence to the committee. "A person checking a decision needs the power to correct it," Cooper said. "One contributor put it to the committee in one line: evaluation without authority produces reassurance, not assurance." Control, not just Australian servers. Two consultancies told the committee what happened on 13 June 2026, when a US Department of Commerce order suspended access to two Anthropic models, Fable 5 and Mythos 5, for users outside the United States. One, The AI Accelerator, wrote that its business continuity depended on the scope of a foreign directive rather than any decision available to it. Access returned on 1 July.
"Keeping data in Australia does not keep control in Australia," Cooper said. "Control is who can switch the service off, who gets the capacity, and whether you can leave."
Who answers when an agent acts? The Humanware Institute recommends settling responsibility for AI agents in payments first, where the consequences are tangible and rules for unauthorised transactions already exist. Those rules address a payment the account holder did not authorise. The report identifies a different case, an agent acting within its permission but beyond the person's intention and finds the application of existing law to it contested or untested. Its comparison of eight other governments did not identify a settled allocation of loss for that case either.
"Agents are already booking, cancelling and paying on people's behalf," Cooper said. "Nine institutions gave evidence to the committee, and none of them told it who carries the loss when your own agent spends within its permission but not as you meant. That can be settled now, before it becomes a dispute, and payments is the place to start."
Delivery and pace The report maps relevant responsibilities across at least 19 Commonwealth ministers and agencies, alongside Parliament and the states. It calls for one named point of accountability for coordination, pace and reporting across existing institutions, not a new regulator, with properly resourced regulators, credible independent assurance, and defined responsibilities, deadlines and measures of progress. The Institute has identified the need for a delivery roadmap that connects Australia's AI activity with responsible commercialisation that can generate measurable improvements in productivity, service quality and working lives. Since returning to Australia in 2016, Cooper has strongly advocated for turning technological innovation into practical economic value. She has spent the past decade working deeply across financial services, where she has focused on the capability, organisational change and governance needed to realise that value.
"Commercialisation, capability and governance need to develop together," Cooper said. "Australia has the talent and ingenuity to lead in valuable applications of AI. Realising that opportunity means helping organisations deploy it responsibly, measuring the improvements it delivers and, critically, ensuring the benefits are shared."
Consultation, parliamentary scrutiny and pilots provide essential foundations for responsible deployment. A delivery roadmap would connect that work with clear milestones, evaluation and decisions about wider use.
"Australia has plenty under way. What it lacks is pace, and someone to set it," Cooper said. "Legislation provides the foundations. It needs institutions that can test systems, learn from incidents, enforce the rules and update them as the technology changes."
The committee's final report is due by 30 November. The report sets out seven questions it will use to assess it.
"Parliament received more than 6,000 pages. We read all of it analysed the full body of evidence covered by this edition, and we are publishing now, so the next person can do the same for any consultation in the country," Cooper said.
