Industry and Innovation Minister Tim Ayres released the federal government’s long-awaited National AI Plan in Canberra on 2 December 2025, anchoring it with a new Australian AI Safety Institute backed by $29.9 million and slated to open in early 2026.
The plan, developed inside the Department of Industry, Science and Resources since 2024, sets three goals: draw investment into Australia’s digital and physical infrastructure, lift AI skills and adoption across the economy, and keep Australians safe as the technology spreads.
Notably, it does so without the mandatory guardrails for high-risk AI that earlier consultations had flagged. The government has instead opted to lean on Australia’s existing laws, a choice that has drawn both relief and criticism.
What the plan actually commits to
The centrepiece is the AI Safety Institute. According to the Department of Industry, Science and Resources, the $29.9 million body will monitor, test and share information on emerging AI capabilities, risks and harms, and support agencies and regulators to respond in time.
“The National AI Plan is about making sure technology serves Australians, not the other way around,” Ayres said in the government’s release, as reported by SBS News. He added that the plan aimed at “capturing the economic opportunities of AI, sharing the benefits broadly, and keeping Australians safe”.
Beyond the institute, the plan covers investment in data centres and compute, faster AI take-up in public services, expanded access to AI skills, workplace protections, and deeper international engagement on AI standards. Ayres framed it as a living document, saying the government would “continue to refine and strengthen this plan” as the technology evolves.
The economic stakes cited are large. SBS noted the Australian Academy of Technological Sciences and Engineering had warned in November that adequate AI investment could deliver a $150 billion boost, while the Productivity Commission has pointed to a potential “$100-billion-plus” gain if heavy regulation does not constrain development.
The regulatory pivot that split opinion
The most consequential decision is what the plan leaves out. As analysed by the International Association of Privacy Professionals, the government has shifted from the mandatory-guardrails model canvassed in earlier consultations to a light-touch approach that will “continue to build on Australia’s robust existing legal and regulatory frameworks”.
Earlier proposals had floated risk management plans, pre- and post-deployment testing, complaint mechanisms, incident reporting and independent audits for high-risk settings. According to consultancy Privacy108, those binding obligations were dropped in favour of existing technology-neutral laws spanning consumer protection, privacy, discrimination and online safety, plus voluntary guidance.
The Productivity Commission had recommended deferring major AI-specific rules, a position that shaped the outcome. Supporters see a pragmatic bet on growth. Critics see a gap.
Privacy108 reports that Electronic Frontiers Australia criticised the approach as prioritising “economic opportunity over citizen safety and digital rights”, and noted survey findings that more than three-quarters of Australians support explicit AI regulation. The IAPP’s analysis flagged concerns that non-AI-specific rules may prove ineffective against risks such as opaque automated decision-making, algorithmic discrimination and unclear accountability across AI supply chains.
Workers, universities and creators weigh in
The response has not split neatly along predictable lines. The Australian Council of Trade Unions welcomed the plan while pressing for enforcement of existing law against large technology firms.
“The Albanese Government’s National AI Plan has workers’ rights at its heart,” ACTU Assistant Secretary Joseph Mitchell said in the union’s official statement. Mitchell argued that “workers aren’t afraid of AI but are rightly sceptical about letting it go unchecked”, pointing to job displacement and surveillance across warehousing, finance and transport.
The ACTU also raised copyright, arguing generative AI has been trained on creative work “without consent or compensation” for journalists, artists and musicians. On that front the government moved separately, confirming it will not introduce a text and data mining exception to Australian copyright law, a decision welcomed by creators.
Research universities backed the roadmap. The IAPP noted support from Group of Eight chief executive Vicki Thomson, whose members praised the plan’s breadth and its potential to strengthen the local AI ecosystem. Privacy Commissioner Carly Kind was also cited among the officials engaging on the plan’s privacy dimensions.
Why it matters
For Australia, the plan is a wager on sovereign capability. The bet is that lighter regulation plus a well-resourced institute will attract data-centre and compute investment, seed high-value local jobs, and let firms adopt AI faster than rivals weighed down by heavier rules.
The risk sits on the other side of that ledger. Without mandatory guardrails, workers, consumers and creators are relying on laws written before generative AI to catch algorithmic harm, workplace surveillance and content scraped without payment. The ACTU’s central ask, that the Safety Institute verify compliance with Australian law, will test whether a $29.9 million body can hold multinational developers to account.
Skills and equity are the other unresolved questions. The plan promises to widen AI training and protect regional and disadvantaged communities from being left behind, but the detail of delivery is still to come.
The forward look
The immediate milestone is the AI Safety Institute standing up in early 2026, when its testing and monitoring remit will start to take shape and its independence will be scrutinised. The government has signalled incremental reform rather than a single AI Act, meaning the real regulatory story will play out through sector-specific updates and voluntary frameworks over the coming year.
Ayres has left the door open, describing the plan as one the government will keep refining. Whether that flexibility reassures investors or worries safety advocates will depend on how quickly Canberra acts when the first serious AI harms land on Australian soil.
Sources: Minister for Industry, Science and Resources media release; SBS News; International Association of Privacy Professionals; Australian Council of Trade Unions; Privacy108.
