Australia's AI Crossroads: Investment, Regulation, and the Search for Balance

The Australian government is actively shaping its approach to artificial intelligence (AI), balancing the attraction of major technology investment with calls for robust regulation on copyright, workplace safety, and national security. This complex landscape involves multiple government departments, international technology firms, domestic political figures, and expert advisors, each presenting distinct perspectives on the path forward.

Government Policy and Regulatory Approach

Shift from an AI Act to a National Plan

The Australian government has moved away from plans for a standalone AI Act, which was previously proposed by former Industry Minister Ed Husic. Instead, in December 2025, the government released a National AI Plan that emphasizes the application of existing laws, such as privacy protections, to manage AI risks.

• Conflicting Statements on Rationale: Ed Husic has claimed that the government abandoned mandatory guardrails on high-risk AI due to pressure from the administration of former US President Donald Trump. Current Minister for Industry and Science Tim Ayres has disputed this interpretation, asserting the policy shift was unrelated to the Trump administration.
• National AI Plan: The plan aims to attract big tech investment and mitigate AI risks.

"The government abandoned mandatory guardrails on high-risk AI due to pressure from the former Trump administration," former Minister Ed Husic claimed.

Establishment of the Australian AI Safety Institute

The government has established the Australian AI Safety Institute, a new body within the Department of Industry, Science and Resources.

• Leadership: Dr. Kate Conroy has been appointed as the inaugural general manager of the institute.
• Funding and Timeline: The institute has secured funding of $29.9 million over four years and is expected to become operational in early 2026.
• Responsibilities: The institute is tasked with monitoring, testing, and sharing information on AI capabilities, risks, and harms. It will address both "upstream" risks (capabilities, building, training of AI models) and "downstream" harms (real-world effects on people).
• International Collaboration: Australia has signed a memorandum of understanding with the UK's AI Security Institute for information sharing and joint research.

Disbandment of AI Advisory Body

The federal government has discontinued its planned permanent AI advisory body.

• Timeline: The project took 15 months and cost approximately $188,000 to narrow a field of 270 experts to 12 nominees.
• Rationale: A spokesperson for Minister Ayres stated that the new AI Safety Institute would offer a more dynamic approach, capable of testing, monitoring, and advising on regulatory gaps, reducing sole reliance on external expertise.
• Expert Reaction: Professor Toby Walsh, a member of the scrapped expert group, expressed concerns that Australia is missing a narrow window to regulate AI effectively.

"Australia is missing a narrow window to regulate AI effectively," warned Professor Toby Walsh, a member of the now-disbanded expert group.

Copyright and AI Training

Current Legal Framework

Current Australian copyright law requires AI companies to negotiate licenses for using copyrighted material.

• Previous Government Decisions: In October 2023, Attorney-General Michelle Rowland declared the government would not grant a copyright exemption for AI training. This stance was reiterated in May 2024. The government also rejected a text and data mining exemption in the previous year, which would have allowed AI companies to use Australian creative works for training without permission.
• Current Stance: The government has repeatedly stated since October 2023 that it will not weaken copyright protections for AI.

Debate on Potential Policy Changes

A debate exists regarding potential changes to copyright rules in exchange for data center investment.

• Senate Exchange: Independent Senator David Pocock alleged in Parliament that a cabinet submission proposes allowing AI companies to use Australian copyright material for training in exchange for data center investment and a creative fund. Minister Tim Ayres called the claim "reckless speculation" and stated the government has not changed its position, ruling out a text and data mining exception.
• Considered Models: Two models are reportedly under consideration: a text and data mining exception (carve-out) or an expanded licensing system.
• Stakeholder Positions:
  • Creative Sector: ARIA chief Annabelle Herd and music royalty body APRA AMCOS have expressed concern over any reversal of policy. APRA AMCOS Dean Ormston stated a copyright exemption would result in "theft of people's creative work."
  • Tech Industry: The Tech Council of Australia and Atlassian co-founder Scott Farquhar continue to call for copyright reform to attract AI investment.
  • Legal Experts: Kimberlee Weatherall and Kathy Bowrey raised questions about payment distribution, creator consent, and the necessity of tying copyright changes to data centers.

"A copyright exemption would result in the theft of people's creative work," argued APRA AMCOS Dean Ormston, opposing any policy reversal.

AI Safety, Cybersecurity, and National Security

Joint Five Eyes Warning

Signals intelligence agencies from Australia, the US, UK, New Zealand, and Canada issued a joint statement warning that frontier AI models capable of enabling devastating cyber attacks on governments and businesses are anticipated within months. The statement says AI will accelerate the speed, scale, and sophistication of cyber threats.

US Export Control Orders

Anthropic received an export control directive from the Trump administration to suspend access for foreign nationals to its Fable 5 and Mythos 5 AI models. Anthropic took both models offline for all customers to comply.

• Anthropic's Response: The company stated it believes there was a misunderstanding and is working to restore access. Access to all other Anthropic models is not affected.
• Contradictory Access Reports: Despite the order, reports indicate that Anthropic has maintained access to a "Mythos Preview" model for early testers, including businesses such as banks and technology firms, through a program called "Project Glasswing."
• Australian Access: Anthropic had recently granted Australian companies and individuals access to these models. OpenAI has begun rolling out a comparable model, GPT-5.5-Cyber, to Australian organizations including Commonwealth Bank and Westpac.

Workplace Safety and Surveillance

A report from the John Curtin Research Centre warned that unregulated AI in workplaces could lead to increased worker surveillance, unsafe workloads, and job insecurity. Recommendations include a national AI taskforce, a review of the Fair Work Act, mandatory human oversight of AI, and an AI expert advisory panel within the Fair Work Commission.

• Government Response: Workplace Relations Minister Amanda Rishworth announced a government forum with employers and unions to discuss AI adoption themes: trust, capability, transparency, safety, and productivity.
• Legal Context: Workplace relations lawyer Shannon Chapman noted Australia's legal framework for AI in workplaces is complex, with no overarching national legislation. Laws vary by jurisdiction and depend on data type and usage.

AI Infrastructure and Data Centers

Government Expectations for Data Centers

The Labor government introduced new national expectations for data centers and AI infrastructure projects. Projects demonstrating economic, green energy, and national interest benefits will receive priority for approval. Industry Minister Tim Ayres stated these expectations aim to prevent a "race to the bottom" regarding water and electricity consumption.

Data Center Development and Local Opposition

Reports indicate significant planned data center development in Australia, with more than $150 billion in planned investment.

• Energy and Environmental Concerns: A report concluded that energy demand from AI data centers is jeopardizing the country's energy transition. Anthropic has expressed interest in building data centers requiring up to 20 GW of electricity by 2030, approximately 60% of Australia's current generation output. Former chief scientist Alan Finkel characterized this demand as "huge."
• Political Opposition: Politicians across the political spectrum have voiced concerns. Senator Sarah Hanson-Young (Greens) called for a moratorium on building and approving new data centres until regulations are established. Senator David Pocock argued that these developments can put pressure on power and water supplies. Former minister Ed Husic warned that Australia is sleepwalking into an AI catastrophe, citing low public trust and lack of community input on data center construction.
• International Context: Opposition to data centre developments is also growing in the US and UK due to concerns about energy consumption and environmental impact. Reports indicate local opposition delayed or blocked projects worth up to $152 billion in the US in 2025.

"Australia is sleepwalking into an AI catastrophe," warned former minister Ed Husic, citing low public trust and a lack of community input on data center construction.

International Comparison

Professor Toby Walsh highlighted that South Korea, Japan, China, Taiwan, and Sweden have introduced comprehensive AI regulation laws. He also noted that Canada has invested six times more in AI over the past five years, and Singapore, with a smaller population, has invested 15 times more.

Industry and International Engagement

Anthropic's Engagement with Australia

Anthropic CEO Dario Amodei visited Australia to meet with Prime Minister Anthony Albanese and Treasurer Jim Chalmers.

• Memorandum of Understanding: In April, Industry Minister Tim Ayres signed a non-binding memorandum of understanding with Anthropic outlining a "shared vision." The company has committed to the government's AI plan, including investing in renewable energy for data centers.
• Office Opening: Anthropic opened a Sydney office in July, its fourth in the Asia-Pacific.
• CEO Statements on Copyright: Mr. Amodei stated that Anthropic is not attempting to alter Australia's copyright protections, acknowledging that rights holders have "legitimate claims." He suggested that if AI generates significant economic growth, the focus should be on increasing overall prosperity.

"Anthropic is not attempting to alter Australia's copyright protections," CEO Dario Amodei stated, acknowledging that rights holders have 'legitimate claims.'

OpenAI's Activities

OpenAI received a subpoena from New York's attorney general seeking documents related to user impact, advertising, and internal policies. The company has also begun rolling out its GPT-5.5-Cyber model to Australian customers.