Half of Australians Unsure When AI Is Safe to Use, Study Finds

A significant digital divide has emerged between the rapid rollout of artificial intelligence and public confidence in actually using it, according to new research published by Good Things Australia ahead of Get Online Week.

The findings show that only 27% of Australians feel confident working with AI tools, with confidence dropping to just 10% for individuals aged 75 and over. Crucially for employers and service providers, 51% of respondents said they are unsure when it is appropriate to use AI tools in day-to-day settings, while 39% expressed concern about the broader impact of the technology on society and the environment.

At the same time, appetite to understand the technology remains strong. More than a third of Australians stated they want to learn more about AI, specifically naming prompt craft, digital safety, practical applications, and the ability to detect and verify machine-generated output as their main areas of interest. In response, community organisations running digital skills sessions across the country from 19 to 25 October are dedicating around half of their programs directly to AI literacy.

Good Things Australia chief executive Jess Wilson noted that simply handing people access to AI software does not automatically teach them how to evaluate outputs, question errors, or handle data responsibly. Closing that divide requires structured, supportive spaces where learners can ask basic questions and build practical habits without fear of making mistakes.

What this means for Australian businesses

For small and medium enterprises, this data provides an important reality check: assuming your staff or customers are comfortable navigating AI on their own is a risky bet.

• Establish practical workplace guardrails: Because more than half of people are unsure when using AI is appropriate, businesses should not rely on unspoken common sense. Create clear, concise guidelines that show staff exactly which tools are approved, where customer data must never be pasted, and when human review is strictly mandatory.
• Upskill beyond tool access: Providing subscriptions to software like Copilot or ChatGPT will not deliver productivity gains if team members lack confidence. Invest time in internal training that focuses on basic prompting, fact-checking outputs, and identifying inaccuracies.
• Support customer comfort: If your business is introducing AI-driven chatbots or automated workflows for clients, remember that many customers remain cautious or sceptical. Always disclose when automation is in use and maintain a frictionless way for clients to reach a real person.