Australian researchers are using AI for coding, literature searches and specialised scientific applications, alongside generative AI tools, but widespread use is accompanied by concerns about its implications for research.
The first results from Australia's National AI in Research Survey highlight concerns about copyright, environmental and community impacts, the power of leading AI firms and reliance on commercial models.
Science & Technology Australia (STA) is releasing the findings as it gives evidence to the Joint Select Committee on Artificial Intelligence this afternoon.
Conducted in partnership with MetaMelb at the University of Melbourne, the survey collected 1,034 complete responses across disciplines and career stages.
STA CEO Ryan Winn said nearly three-quarters of respondents reported using AI but the findings reinforced the urgent need for better support.
"AI can accelerate discovery and lift productivity. But researchers also need the skills, computing power, secure data and expert support to use it well," Mr Winn said.
"They see opportunities to accelerate discovery, alongside significant concerns about how AI is developed and used. These findings underline the need for clear guidance, appropriate safeguards and the skills to assess when AI is useful and when it introduces risks."
More than half of respondents (54 per cent) considered their organisation's AI training and skills-building opportunities inadequate. Institutional guidance was frequently described as unclear or poorly coordinated.
"Researchers need support to evaluate AI tools, verify their outputs and protect sensitive information. That is essential to realising the benefits while maintaining research integrity," Mr Winn said.
The findings also reveal clear boundaries around acceptable use. Researchers generally accepted AI assistance with drafting and editing grant applications but considered using AI to determine funding scores or assess individual researcher or institutional research performance unacceptable.
"Researchers see value in AI assistance but draw a firm line at handing over scholarly judgement. Also, decisions about research funding, careers and performance carry real consequences. People must remain responsible for them," Mr Winn said.
Computing access is another critical challenge, with security and privacy concerns the leading reported barrier to using AI-related computing resources.
Fewer than one in ten respondents quantified their computing requirements for the next 12 months. However, their estimates totalled approximately 53.5 million CPU-hours and 14.3 million GPU-hours.
"These figures are not a specific national demand forecast, but they do reinforce the need for coordinated planning and investment," Mr Winn said.
"Applications for Australia's 2026 national research computing allocations sought more than 2.2 billion hours, almost three times the combined capacity of our two national public facilities."
STA welcomed the Committee's recommendation last week for continued investment in public research computing through programs like the National Collaborative Research Infrastructure Strategy. This should be the foundation that any private capacity should complement, not replace.
Mr Winn said these findings should inform the 2026 National Research Infrastructure Roadmap, whose exposure draft is currently open for consultation.
"The Roadmap is an opportunity to plan for how AI is changing research across every discipline. Computing power, secure data, storage and the specialist workforce must be planned and funded together," he said.
"Commercial contributions can expand access, alongside sustained public investment in facilities such as the National Computational Infrastructure and Pawsey Supercomputing Centre."
The results also reinforce STA's calls for the Government to implement the Ambitious Australia report's recommendations to strengthen Australia's research and development system.
"These findings give the Government timely evidence to guide the Roadmap and implementation of Ambitious Australia, and turn Australia's AI ambitions into lasting public benefit."
