Australians wary of AI spending, identity data trust

Australian consumers would let an AI agent spend an average of AUD $200 on a routine purchase without final approval, according to research from Ping Identity. However, 53% would cap that spending at AUD $143 or less.

The findings come from a study of 1,000 Australian consumers within a wider global survey of 11,000 people across 12 countries. In Australia, the results suggest a cautious view of AI when it moves beyond answering questions and starts acting on a user's behalf.

Only 18% of Australian respondents said they were likely to let AI independently make decisions for them, while 38% were not comfortable giving AI any of the levels of responsibility tested. Globally, 25% rejected all responsibility levels in the survey.

Australians also placed a high value on disclosure, with 68% saying they wanted to know whenever AI was being used, the highest share among the countries surveyed.

Trust gap

The research also found low confidence in how organisations handle identity data. Just 7% of Australian consumers said they fully trusted organisations to manage their identity data.

Concerns also extended to privacy and security. Some 42% of Australian respondents worried AI would increase cybersecurity risks to their personal data, while 43% worried it would invade their privacy.

Government oversight was another priority, with 83% of Australian consumers saying regulation to protect identity data in AI applications was important.

Awareness of newer forms of AI also appeared limited. Some 67% of respondents said they were either not very aware or not aware at all of what agentic AI means.

Email and autonomy

Consumers were particularly reluctant to grant access to core personal services. Only 21% said they would be likely to give an AI agent read access to their primary email, even if it saved them five hours a week.

Representation and supervision were also key concerns. Some 77% of Australians worried an AI agent acting on their behalf could misrepresent their personality or personal values.

In addition, 43% said they were more concerned about having to constantly double-check an AI agent's work than about the agent making a mistake, cited by 31%.

"Consumers are open to AI doing more, but they want a say in how far it can go," said Darryl Jones, Vice President of Consumer Segment Strategy at Ping Identity.

"The more responsibility we give AI, the more important it becomes to be clear about what it can do, what it can access and who remains in control," Jones said.

Global comparison

Globally, people were more comfortable with AI in support roles than in autonomous ones. Ping Identity found that 56% of consumers were comfortable using AI for low-stakes support or recommendations.

That comfort dropped sharply when action was involved. Just 5% of global respondents said they were comfortable with AI taking action with approval, while only 3% were comfortable with routine actions or transactions without final approval.

Attitudes also varied significantly between countries. Consumers in India and Indonesia were much more open to letting AI independently make decisions on their behalf, at 70% and 61% respectively. The figures were 21% in Sweden, 19% in the Netherlands and 18% in Australia.

Views on disclosure also differed. Australia, the US and the UK showed strong demand to know when AI was being used, while Indonesia stood apart, with 62% of consumers saying speed and accuracy mattered more than disclosure.

The global data also pointed to similar concerns about representation and oversight. Across all surveyed markets, 71% worried that AI agents acting on their behalf could misrepresent their personality or values, while 47% said the need to keep checking an AI agent's work was a bigger concern than the agent making a mistake.

Trust in organisations to manage identity data has also weakened globally. Only 12% of respondents worldwide said they fully trusted organisations to manage their identity data, down from 17% in the previous year.

Talker Research conducted the study on behalf of Ping Identity using a random double-opt-in survey model. It covered consumers in the US, UK, France, Germany, Australia, Singapore, India, Indonesia, the Netherlands, Sweden, Spain and Italy.