Australian parliamentary inquiry leaves liability for AI agent purchases unresolved

A study of evidence submitted to the Australian Parliament's Joint Select Committee on Artificial Intelligence shows widespread backing for automated technology, but reveals a legal blind spot around financial accountability for AI agents.

The review was conducted by the Sydney-based Humanware Institute and released on 5 November 2026. Titled 'AI on Australia's Terms: What Parliament's AI Inquiry Reveals About Power, Value and Accountability', the 106-page paper analysed public testimony from 575 participants, covering 564 written submissions and 53 hours of public hearings.

According to the analysis, nine organisations in the financial system gave evidence to the committee, including three major Australian retail banks, their industry association, Treasury, and four economic regulators. When asked who carries the financial loss if an AI agent carries out a transaction within its technical permissions but beyond what a person intended, none of the organisations could provide a clear answer.

Banks suggested that software makers who develop the agents should bear responsibility. In contrast, the consumer regulator pointed to Treasury, the securities regulator stated existing laws apply, and the Productivity Commission referred to pending reviews. Current financial regulations govern completely unauthorised transactions, such as stolen cards, but remain untested for autonomous agents authorised by an account holder that later make incorrect buying decisions.

Agentic transactions are already underway locally. Mastercard recorded the first authenticated agent-led card purchases in Australia in January 2026 during bank pilot programs. The Reserve Bank of Australia is reviewing payments system rules for agentic commerce, but recommendations from that work are not due until 2028.

The analysis found that 367 out of 575 participants supported AI adoption, though 290 of them asked for safeguards, worker consultation, and clear accountability rules.

What this means for Australian businesses

For small and medium enterprises exploring autonomous AI agents to book supplies, manage inventory, or handle commercial purchases, the lack of legal clarity carries operational risk:

• Set strict budget caps and transaction thresholds inside any software that uses automated buying or booking capabilities.
• Require human sign-off for purchase orders above minimal amounts rather than giving software agents complete autonomy.
• Keep detailed activity logs for agent tools, recording the precise prompts, approvals, and API actions that trigger commercial transactions.
• Clarify commercial terms and chargeback liabilities with payment providers and business credit card issuers before connecting cards to agentic software.