The disruptive arrival of Chinese artificial intelligence chatbot DeepSeek presents new opportunities for Australian technology companies but may also attract concerns around privacy and national security, experts say.
DeepSeek, which rose to the top of app store charts and wiped $1 trillion in value off US tech stocks this week, has been deemed a game-changer by many industry figures after the company claimed to have developed a powerful large language model (LLM) at a fraction of the cost and computing power of America’s largest AI labs.
The revelation, which some have heralded as a breakthrough, has spooked Silicon Valley investors and presented the world with a new open-source model which can be used to improve other systems, can be checked for security gaps, and can even run offline on a local device.
DeepSeek’s R1 model could enable greater diversity, accessibility, and innovation in AI, decrease reliance on financial and technological capital, and create a more level playing field for smaller players such as Australia, experts said.
DeepSeek suggests Australia is ‘not too late to the party’
Professor Anton van den Hengel, chief scientist of the Australian Institute for Machine Learning at the University of Adelaide, said DeepSeek’s rise meant Australia was “not too late to the party” when it came to AI investment and development.
“DeepSeek means it’s not out of reach — we can do it — and it’s an illustration of the fact there are plenty of opportunities left for high-tech, well-engineered breakthroughs to disrupt the hegemony,” he told Information Age.
Australia’s newly appointed chief scientist, Tony Haymet, told a Tuesday press conference that DeepSeek made a "genuine" impact on consumers and technology markets which highlighted “how disruptive technology can be, and how quickly things can happen”.
Haymet said he saw AI as “a great opportunity” for Australia, given the nation was “perfectly set up” to use renewable energy to run power-hungry AI data centres — a sentiment echoed by van den Hengel.
“We could be an information processing centre for the rest of the world," he said.
"… The wonderful thing about information processing is that it’s a lot cheaper to ship bits than electrons.
“DeepSeek is a pointer to the fact that Australia can become an information processing hub globally — because the chips don’t matter anymore.”
DeepSeek stunned Silicon Valley by purportedly producing its R1 model at much lower cost and with less computing power than US competitors. Photo: Shutterstock
Asked about DeepSeek and the Australian government’s work on AI, Minister for Industry and Science Ed Husic said while he thought Australians would “naturally gravitate” towards being cautious of DeepSeek as some have been of Chinese social media platform TikTok, he saw DeepSeek as “the latest chapter” in a race which Australia would “have to get accustomed to”.
“The big thing is to release the handbrake — and that handbrake is constituted by a lack of confidence and trust in technology, and I think we’ve got to work on that,” he said on Tuesday.
The US still led the world in AI development and Australia would continue to benefit from its research ties to America, Husic said.
The Tech Council of Australia, which represents several companies with large AI businesses —including Microsoft, Adobe, Google, IBM, and Amazon Web Services — said it urged the government to “act now or risk Australia falling behind in AI development and adoption”.
The industry group said while it supported the government’s work to develop a National AI Capability Plan, it argued Australia “cannot wait for the plan to be finalised by the end of 2025” and urged the scheme and any related investments to be brought forward.
Husic’s office said the minister held roundtable discussions with “leading corporate users of AI and unions” on Wednesday to consult on the capability plan.
Industry and Science Minister Ed Husic and Australia's new chief scientist Tony Haymet responded to questions about DeepSeek on Tuesday. Image: Australian Parliament / YouTube
Where are the Australian LLMs?
Professor van den Hengel said DeepSeek’s rise showed it was “more important than ever” for Australia to develop its own sovereign AI capabilities, including a domestic LLM.
“We definitely still need to build our own Large Language Model, but the benefit you get out of building a LLM isn’t to have a LLM — that’s a subsidiary benefit, it’s fantastic — but the real benefit is that you train people in the skills of building LLMs.
“So whatever the next thing is, they can pivot quickly to do that,” he said.
“High-tech skills in the economy is what we need to build as sovereign capability … in order to participate in the modern AI-enabled economy.”
While several LLMs are in development in Australia, none are powerful chatbots available for general consumer use.
Kangaroo LLM, a local AI project staffed largely by unpaid volunteers and supported by a consortium of technology companies, launched in 2024 with the goal of creating Australia’s first open-source LLM built with local language and culture in mind.
The project has not provided a substantial update on its progress in months, and did not respond to a request for comment.
Van den Hengel said while undertakings such as Kangaroo LLM were “fantastic” for building enthusiasm, they were “not the final solution”.
“We need to engage with this challenge at a much greater scale, as a nation, if we’re going to solve that problem,” he said.
Van den Hengel pointed to the success of Australia’s Leonardo.AI and its image-generating model Phoenix, which was released publicly in June 2024 before the company was acquired by Australian tech giant Canva.
Australian organisations have also developed some industry-specific LLMs such as systems trained on legal documents for use by legal professionals, and the ABC has also been developing its own internal LLM.
On Tuesday, Husic teased funding announcements for Australian AI firms to be made “in due course” and cited $1 billion of available critical technology funding in the National Reconstruction Fund (NRF), which would include some AI projects.
The NRF announced a $32 million stake in Australian AI health company Harrison.ai on Wednesday.
DeepSeek’s China links ‘could be troubling’
DeepSeek's rise to prominence has drawn some concerns from the government and cybersecurity experts, due to obligations on Chinese companies to comply with data requests from the Chinese government.
Senior cybersecurity and networking lecturer Dr Saeed Rehman from Flinders University said while DeepSeek’s terms and condition were “similar to those of other AI providers”, its data collection “would raise significant concerns”.
This included user inputs being accessible to humans for training purposes, being stored on servers in China, and containing information such as user locations and “keystroke patterns or rhythms”, according to DeepSeek’s privacy policy.
“The fact that user data is stored on servers in China, a country known for its stringent data control policies, could be troubling for users and governments wary of their data privacy,” Rehman said.
“This situation may evoke similar concerns to those raised for TikTok, where data privacy and security have been hotly debated and led to bans in some Western countries."
The Australian government has previously said it has no plans to ban TikTok.
DeepSeek has also displayed political biases from its Chinese training data, such as refusing to provide details about the 1989 Tiananmen Square massacre.
Asked about potential national security implications for Australians using DeepSeek, Husic said it was “too early to jump to conclusions” such as banning the app domestically, but added the government was ready to act if it believed it needed to.
The minister also questioned whether DeepSeek’s approach to data and privacy would be welcomed by Australians, who he argued were often wary of AI.
On Wednesday, Treasurer Jim Chalmers added the government “would urge Australians to be cautious about this new [DeepSeek] technology”.