Australia’s ARC and NHMRC Allowed Limited Reviewer Use of Generative AI from April 2026

On 28 April 2026 the Australian Research Council (ARC) and the National Health and Medical Research Council (NHMRC), Australia’s two main national research funders, released updated policies on the use of generative AI by applicants and assessors, together with a joint statement by their chief executives. The ARC policy, version 2026.1, replaces the ARC’s 2023 policy. NHMRC revised its own policy on the same day.

The central change concerns peer review. The ARC’s 2023 policy, issued on 7 July 2023, stated that peer reviewers, including all detailed and general assessors, must not use generative AI as part of their assessment activities, and that suspected misuse would lead the ARC to remove the assessment. NHMRC’s earlier guidance similarly stated that peer reviewers must not input any part of a grant application into an AI system to assist in assessment.

Under the 2026 policies, the ARC permits assessors to use generative AI tools provided that conditions are met, including preserving confidentiality. Peer reviewers may use AI to refine the wording of review comments, for example to improve grammar, but must rely on their own expertise when evaluating and scoring applications. Assessors may not use AI to form their judgement of an application’s quality. NHMRC’s policy lists permitted uses such as checking grammar or reducing language barriers.

The ARC’s 2023 policy had also treated uploading application material to generative AI tools as a breach of confidentiality. The 2026 policy keeps confidentiality as a condition of any permitted use, so the change is one of degree: limited assistance with expression is allowed, but release of confidential content and delegation of judgement are not.

Both funders allow applicants to use generative AI as a support tool. Responsibility for the accuracy and integrity of any AI-generated output remains with the applicant. This is consistent with the ARC’s 2023 approach, which advised caution for applicants rather than prohibition.

Secondary sources report that the 2026 policies encourage transparency about AI use where it adds value but do not mandate specific disclosure statements. One third-party tool still describes mandatory disclosure, which appears to reflect earlier guidance. Applicants should check the current policy documents and scheme-specific grant guidelines for any declaration requirement.

The ARC policy applies to Discovery Indigenous 2027 and later scheme rounds. The NHMRC policy applies to all peer review conducted for NHMRC MREA grant schemes from 28 April 2026.

The Australian change places the ARC and NHMRC among funders that have moved from outright bans to conditional use. In the United Kingdom, a revised joint statement by UKRI, Wellcome and other funders in June 2026 also narrowed the reviewer rule to a prohibition on inputting confidential content into generative AI tools. By contrast, the US National Institutes of Health continues to prohibit its peer reviewers from using generative AI to analyse applications or formulate critiques under notice NOT-OD-23-149.

For researchers who apply to funders in several countries, these differences mean that a practice permitted for an ARC or NHMRC review, such as using AI to polish the grammar of comments, may still be prohibited for an NIH review. Reviewers should check the rules of each funder separately.

The revised AI policies arrive while the wider Australian research integrity system is itself under examination. On 23 February 2026 the ARC and NHMRC chief executives said they were working with Universities Australia to understand how the integrity system operates and where it can be strengthened, including clearer institutional responsibilities under the Australian Code for the Responsible Conduct of Research. Misuse of generative AI in applications or reviews could therefore be considered both under the AI policies and under the integrity framework that institutions apply to their staff.

Reviewers should use generative AI, if at all, only for permitted tasks such as refining wording, should use tools that preserve confidentiality, and should form and record their evaluation and scores independently. Applicants should check every AI-assisted passage for accuracy, keep records of how AI was used, and follow scheme guidelines on transparency.

Research offices should update internal guidance and training, and should confirm which tools available within the institution meet the confidentiality conditions. Grant writing consultants and medical writers working on ARC or NHMRC applications should agree AI use with the lead investigator in advance. This brief summarises public ARC and NHMRC documents for information and is not affiliated with either agency.