The Australian Research Council (ARC) has updated its policy governing the use of generative artificial intelligence across its National Competitive Grants Program (NCGP), effective 28 April 2026. The revised policy applies to applications and assessments for Discovery Indigenous 2027 and to every scheme round opening from that date onward, and it was released alongside a joint statement from the ARC and Australia’s National Health and Medical Research Council (NHMRC) signalling a coordinated approach across the country’s two largest competitive research funders.
What the policy actually says
The policy distinguishes clearly between applicants and assessors, and sets different rules for each.
Applicants
Applicants may continue to use generative AI tools as a support tool while preparing a proposal — the ARC’s guidance specifically names tasks like testing ideas, improving language, and summarising text as acceptable uses. Critically, however, the policy does not shift authorship or accountability: applicants and the researchers named on an application remain fully responsible for its intellectual content, and must ensure nothing submitted is false or misleading. Administering organisations (the universities and research institutions submitting on researchers’ behalf) must have their delegated research leader certify that named participants are responsible for the authorship and intellectual content of the application and have not infringed anyone’s intellectual property rights in preparing it.
Assessors and peer reviewers
The policy is considerably more restrictive for the assessment side of the process. Assessors may use generative AI only for narrow, mechanical purposes — correcting or improving the grammar, spelling, formatting, and readability of an assessment they have already drafted themselves. They are explicitly barred from using AI tools to help form an opinion on the quality, merit, or fundability of an application. Confidentiality is treated as a hard line: entering any application material into a public generative AI tool (the ARC’s own guidance names ChatGPT, Gemini, Claude, and Perplexity) is described as a serious breach of confidentiality and is not permitted under any circumstance. A suspected breach can result in an assessor’s review being set aside and referral for institutional consequences.
Why the ARC updated the policy now
The update follows a broader pattern across national funders since 2024–2025: agencies have had to move from blanket AI bans (impossible to enforce and increasingly out of step with how researchers actually draft text) toward policies that separate legitimate drafting assistance from uses that would compromise either authorship integrity or the confidentiality of peer review. The ARC’s framing — permissive for applicants preparing their own material, restrictive for assessors handling confidential third-party material — mirrors the shape of policies already adopted by journal publishers and other funders, though the specific rules and enforcement mechanisms are the ARC’s own.
Alignment with NHMRC
The ARC released its update alongside a joint CEO statement with NHMRC, Australia’s medical and health research funder, confirming that NHMRC’s own updated AI policy follows the same underlying logic. For institutions that hold both ARC and NHMRC funding — the majority of Australia’s research-intensive universities — this alignment matters operationally: research offices do not need to maintain two materially different AI-use policies for the two funders’ National Competitive Grants Program and Medical Research Endowment Account schemes respectively.
What this means for research offices
For research administrators supporting ARC applications, the practical implications are: update any internal AI-use guidance to reflect the applicant/assessor distinction rather than a single blanket rule; make sure delegated research leaders understand the certification they are now formally making about authorship and IP; and treat the confidentiality prohibition on assessors as a standing instruction to anyone at the institution who reviews ARC applications, not just those on formal panels. Because the policy applies from Discovery Indigenous 2027 onward, institutions preparing Discovery Indigenous or other 2027-round applications should confirm current guidance reflects the 28 April 2026 policy rather than any earlier, less specific AI guidance the ARC previously published.
Related CASRAI resources
For a broader comparison of how major funders are handling generative AI in grant writing, see CASRAI’s guide to AI tools for grant writing and the comparison of NIH, NSF, and ERC AI policies for grant writing. For the underlying vocabulary this policy relies on, see the dictionary entries for generative AI, generative-AI disclosure statement, and AI tool disclosure.
