The short version
- Australian defense and intelligence leaders warn that relying on offshore AI infrastructure leaves the nation vulnerable to advanced cyber threats.
- The government seeks to attract major US developers by offering local data hosting, contingent on strict oversight and potential copyright law changes.
- Creative industries and some politicians oppose automatic access to copyrighted material, arguing for licensing deals rather than broad exemptions.
A parliamentary inquiry into artificial intelligence has highlighted a sharp divide between the urgent need for advanced cyber defense capabilities and the political risks of accommodating powerful foreign technology firms. Liberal MP Andrew Hastie opened proceedings by illustrating the current limitations of generative AI tools, recounting an incident where Microsoft Copilot suggested he reply with congratulations to a constituent who had written to announce plans for voluntary assisted dying. This anecdote served as a springboard for broader concerns about the reliability of American-run software and the strategic imperative for Australia to develop sovereign AI infrastructure.
Defense and intelligence officials testified that isolating Australian systems from global AI advancements could prove catastrophic for national security. Abigail Bradshaw, director general of the Australian Signals Directorate, emphasized that her agency relies on cutting-edge models to counter highly capable malicious cyber actors. She warned that if adversaries possess newer, more sophisticated tools, they can identify and exploit network vulnerabilities before domestic defenders can respond. The agency is currently entirely dependent on US-based processing power, a dependency that would severely degrade operational capacity during a conflict.
To mitigate these risks, the government is pursuing a strategy to bring major AI developers, including Anthropic and OpenAI, into Australia. Chris Crozier, the defense department’s chief AI officer, stated that the goal is to keep compute, application, data, and decision-making processes within Australian borders. This approach aims to prevent scenarios where offshore providers could abruptly cut off access due to external pressures or internal decisions. By contracting companies like Google to build interlinked datacenters across the country, the government hopes to establish a resilient, locally controlled technological foundation.
Bringing these firms under Australian jurisdiction offers additional benefits beyond physical infrastructure. Officials argue that local presence would facilitate better integration with technical experts and provide earlier, more detailed access to frontier models. The Office of AI noted that subjecting these companies to domestic law would also empower local law enforcement to combat threats emerging on their platforms. Strict conditions for market entry, such as mandatory government oversight and guaranteed access to advanced capabilities, are being considered to ensure that Australian interests are protected.
However, the path to attracting these tech giants is fraught with legal and political challenges, particularly regarding copyright. AI developers have indicated they will not train models in Australia without significant reforms to intellectual property laws. A proposal to grant companies default access to all online content created by Australians has sparked intense backlash from creative industries. Artists and writers argue that companies should negotiate licensing deals rather than receiving blanket exemptions, fearing that such changes would devalue their work and undermine the cultural economy.
Jenna Priestly, an assistant secretary at the attorney general’s department, explained that developers require access to billions of data points for training, making individual licensing agreements practically impossible. She noted that Australia’s copyright framework protects both local and worldwide content, meaning companies would face immense hurdles in securing permissions from every rights holder. The government is currently consulting on standards that would require AI firms to report breaches and potentially reserve computational resources for local research, but the specifics of copyright reform remain unsettled.
Political opposition to rapid deregulation is mounting. Greens senator David Shoebridge, who advocates for a pause on AI development and opposes copyright changes, reported being excluded from the inquiry committee by Labor ministers. Meanwhile, government officials are engaging directly with executives from OpenAI and Anthropic, including special envoy Jeff Bleich, who suggested that Australia has only months to implement effective regulation. The prime minister’s department aims to draft broader AI standards by the end of 2026, balancing the need for innovation with public safety and economic fairness.
The stakes are high as Australia attempts to navigate this complex landscape. Delaying decisions could result in lost influence over global AI development and reduced access to critical technologies, according to Bradshaw. The government must reconcile the demands of national security agencies for immediate, unrestricted access to advanced tools with the concerns of creatives and citizens regarding data rights and corporate power. As consultations continue, the outcome will shape not only Australia’s technological sovereignty but also its ability to defend against evolving digital threats in an increasingly interconnected world.
Sources behind this briefing
Go to the original reporting
- The Guardian World↗Andrew Hastie says AI advised him to reply ‘congratulations!’ to man who planned to end life with assisted dying
