The Significance of AI in US-China Strategic Competition
The Hon. Dr. Kevin Rudd AC
26th Prime Minister of Australia
Global President of the Asia Society, New York
The Bleich Lecture
The Jeff Bleich Centre for Democracy and Disruptive Technologies
Flinders University, South Australia
Adelaide Convention Centre, Wednesday 16 September
It’s great to be back in this great city of Adelaide. I’ve lost count how many times I’ve been here over the years.
This is my wife Thérèse’s hometown where she has so many fond memories as a girl growing up in Walkerville.
We’ve also spent some wonderful time together on Kangaroo Island which is now known to people across the world as a unique travel destination.
And not that many years ago Therese and I celebrated our wedding anniversary in the Barossa where we chilled for a very long weekend and sampled some of the finest wines in the world.
You are indeed blessed to live in such a great state.
It was good too to meet with Premier Malinauskas this morning to discuss progress on the new submarine construction facility under the AUKUS program at Osborne.
This project will be transformational for the South Australian economy as Adelaide and Perth emerge as critical centres for the future of the Australian defence industrial base. Not just subs. But across the full range of advanced defence technologies where AUKUS has now created an increasingly seamless market between Australia, the US and the UK which Australian companies are now exploiting.
Having just completed my term as Australian ambassador to the US, I’ve spent the better part of the last three years working for the prime minister to ensure that the AUKUS project was legislated through the US Congress and fully embraced by both the Biden and Trump administrations.
Mission accomplished!
At every level, AUKUS is now full speed ahead. It’s now all about implementation. And it is creating critical, sovereign defence capabilities for Australia that will be essential for our long-term national security.
And bringing us to the subject of the day, it’s also good to be here with my friend and colleague Ambassador Jeff Bleich. Jeff served as a distinguished US ambassador to Australia under President Obama. In my prime ministership, Jeff and I worked closely together on deepening the US-Australia relationship at a time of growing contest in the Asia-Pacific region.
The topic I’m addressing today is artificial intelligence and the role it plays in strategic competition between China and the United States.
The AI Challenge
The reality every nation confronts today is that artificial intelligence is the technology that is driving a new industrial revolution sweeping across the world.
Together with the rise and rise of China and the relentless onslaught of climate change, the artificial intelligence revolution is one of the three great, global mega-changes washing over Australia’s shores.
I spoke on these challenges last week at the National Press Club in Canberra. They are challenges that require our urgent national response. Business as usual no longer works.
We were once called the Lucky Country.
We have now become the Complacent Country.
We need a new national mindset that drives us to actively carve out our economic and national security future in what is an increasingly uncertain and dangerous world.
That includes building a sovereign AI industry and capability set in Australia, that maximises inbound investment from the United States across the AI stack so that we can compete in compute, frontier models and in intelligent diffusion across the Australian economy.
Yes, there are downsides. But these can be navigated. In fact, we need a robust AI sector in Australia to be able to navigate these risks, particularly in rolling requirement of crafting and re-crafting our cyber defences.
The uncomfortable truth is that the AI genie is out of the bottle. It can’t be put back. That’s an illusion. Our urgent national task now is to manage it.
But if we miss the AI revolution by not building our own sovereign companies, capabilities and competencies, the brutal reality is that Australia will no longer be counted in the first rank of nations.
Our national mindset, our national psychology must therefore change.
AI and Strategic Competition
Artificial intelligence has not only become the future battle ground for economic competitiveness for corporations and nations.
Because of its military applications, artificial intelligence also looms as a deeply disruptive technology capable of delivering asymmetric advantages in the future battle space.
Anyone doubting this should have a long hard look at what now transpires in the battlefront separating Russian and Ukrainian forces in which AI has transformed drone warfare.
When we look carefully at the future of artificial intelligence in shaping the future of US-China relations, we therefore find ourselves confronting three critical domains:
First, how AI is rapidly transforming the economic competitiveness of Chinese, American, and international firms;
Second, how the race for military domination is unfolding, including the critical question of whether the US and China (and for that matter other states) can continue to preserve a “human-in-the-loop” in the future deployment of weapons systems; and
Third, whether despite the depth and breadth of this economic and strategic competition between the world’s two largest powers, there is sufficient mutual self-interest at stake to compel both Beijing and Washington to consider the development of bilateral guardrails which reduce the risk of artificial intelligence generating uncontrollable dangers for us all.
What I propose to focus on in this lecture is the latter - namely whether there is sufficient consensus developing between Washington and Beijing to at least agree on some de minimis guardrails for the future and, if so, what the content of such guardrails might be.
The US Regulatory Approach to AI
Under the Trump administration, the development of artificial intelligence has been driven by an overwhelmingly deregulatory approach.
Executive Order 14179 of January 2025 -- signed in the immediate wake of President Trump’s inauguration -- was entitled “Removing Barriers to American Leadership in Artificial Intelligence."
This executive order began a process of repealing the previous, more cautious executive orders of the Biden administration on artificial intelligence.
And it directed the administration to remove all federal policies that could be considered obstacles to US global AI leadership.
This was reinforced by the release in July 2025 of "America's AI Action Plan”.
This called for: accelerating innovation; removing burdensome regulation; rapidly expanding domestic AI infrastructure including data centres and energy infrastructure; strengthening export controls on semiconductors; promoting American AI exports; and ensuring federal procurement of AI models was “free from ideological bias”
This deregulatory agenda was reinforced in March 2026 when the US administration released its “National Policy Framework for Artificial Intelligence".
This was a legislative proposal to the US Congress calling for federal pre-emption of any US state AI regulations, in particular by the Californian, Texan, and Illinois state legislatures.
The proposed national framework also sets out to limit developer liability for third party misuse of AI models.
And it seeks to enhance IP protections for AI developers.
President Trump is predictably bullish about the future of American AI:
“We're leading China in AI. We're the most sophisticated country in the world. And frankly, I want to keep it that way, because whoever wins AI, wins. We could put up guardrails. We can do this and that. But I think you have a lot of very negative forces that are bringing this up that shouldn't be bringing this up . And they're bringing up things that won’t happen”.
Treasury Secretary Scott Bessent has reinforced the president's hard line:
“There is no day after tomorrow if China wins at this. If they were to pull away from us at AI, then nothing else would matter… The technical word for stealing and copying American AI models is distillation. So, if the Chinese distil our models, they can never get ahead of us, they can only copy what we already have done.”
Notwithstanding the overall thrust of the administration's deregulatory strategy on artificial intelligence, there are still, however, multiple US government agencies engaged in the question of frontier-AI governance.
These include the: Office of Science and Technology Policy in the White House (OSTP); the Commerce Department’s National Institute of Standards and Technology (NIST) and its Center for AI Standards and Innovation (CAISI); National Security Agency (NSA); Cybersecurity and Infrastructure Security Agency (CISA); Office of the National Cyber Director (ONCD); and National Nuclear Security Administration (NNSA).
The challenge, therefore, faced by the Department of Treasury under Secretary Bessent in coordinating US AI strategy in his engagement with Chinese counterparts is formidable in its complexity.
China’s AI Strategy
China likewise has a bullish strategy on the future of artificial intelligence. Last month, I delivered a lecture to the Australian National University on China's techno-industrial strategy including the ideology of artificial intelligence and Xi Jinping's concept of “New Quality Productive Forces”.
Xi believes that “New Quality Productive Forces”, driven by a series of technological revolutions of which the most important is the artificial intelligence revolution, are now the primary drivers of historical change across the world.
This caused Xi back in 2019 to formally categorise “data” as a fifth factor of production together with land, labour, capital and technology.
For Xi, New Quality Productive Forces driven by AI represent a major quantitative and qualitative leap in the economic transformation process compared with the deployment of the traditional factors of production.
Xi believes that artificial intelligence, augmented by a centrally controlled state, can now move with speed, scale and flexibility to achieve an efficient allocation of resources across the entire economy.
Xi believes this can be achieved through a superior, algorithmic understanding of the market and that this provides the Chinese socialist state with a new and decisive advantage over its capitalist competitors.
According to Chinese official literature, this will produce a revolution in total factor productivity in China, thereby overtaking the traditional productivity dividend thought to uniquely accumulate to the capitalist west.
In China’s view this means that the “visible hand of the Chinese state”, deploying the tools of artificial intelligence within firms across industry sectors – and then across the economy as a whole – can now surpass the traditional efficiencies delivered by Adam Smith's “invisible hand” of the market.
This in turn causes the Chinese system to believe that the rapid deployment of artificial intelligence, as part of a wider set of “new quality productive forces” to which these technologies give rise, is producing a fundamental paradigm shift between China, the US and the West.
They argue that China is no longer in “catch-up” mode with the US. Nor will China simply be in the business of “keeping pace” with the US.
They now argue that through a new “qualitative leap”, driven in large part by artificial intelligence, China is now overtaking the US and the West in productivity growth and, in time, overall economic power. For these reasons, Xi believes that China’s full embrace of artificial intelligence over the last ten years is a fundamental game changer in his country's strategic competition with America.
Meanwhile on the regulatory front, while China has no single AI law, it has committed to developing a single, comprehensive legal code. Nonetheless, the drafting process has been underway for a long period of time. And it has been repeatedly deferred, presumably because of rapidly changing developments in AI technology itself.
Instead, China has developed a dense, fast-moving set of binding rules and industry guidance statements issued vertically by individual government agencies.
These seek to regulate the application of AI in its impact on social stability, state control, as well as the level of AI diffusion they believe is necessary for transformational economic growth.
This regulatory framework includes: first, the Interim Measures for Generative AI Services; second, the AI-Generated Content Labelling Requirements; third, an AI Safety Governance Framework; and, fourth, China’s general Global AI Governance Initiative.
2026 has also seen a new series of regulations concerning AI and the human-machine interface.
These human-machine interface regulations include: the 1 July 2026 TC260 Ethics and Safety Guidelines on personal information and compliance audits; and the 15 July 2026 AI Anthropomorphic Interaction Service Measures (the first dedicated national regime anywhere for AI companions and emotional chatbots which, among other things, mandates addiction prevention and crisis intervention).
In addition to these various regulatory interventions, the Chinese state has also set significant diffusion targets to ensure that AI is maximally deployed across the entire Chinese economy.
This is reflected in the government’s Guidance Opinions on Intelligent Agent Deployment of 15 July 2026 which sets targets for AI agents deployed in the healthcare, transport, media and public safety sectors – affirming a 70% smart terminal adoption target by September 2027.
What we can see from the above is that until now, China’s principal regulatory intervention has focussed on the individual applications of artificial intelligence and their impact on social stability and state control.
Beyond these concerns, the Chinese state has sought to maximise the dissemination of artificial intelligence across the breadth of the Chinese economy through its diffusion targets (i.e. the 70% by 2027 target).
What Chinese AI regulation has not sought to do, however, is to regulate frontier model capability, or any catastrophic risks to safety and security that might arise from such models.
The bottom line is that China's regulatory approach to AI is in large part an industrial policy predicated on winning the economic competition race against the United States.
But it is an industrial policy with a robust censorship spine.
China, like the United States, also faces significant coordination problems across the various agencies of the Chinese state in developing a coherent artificial intelligence strategy and regulatory framework.
For China, the most relevant agencies include: the Cyberspace Administration of China (CAC); Ministry of Industry and Information Technology (MIIT); Ministry of Science and Technology (MOST); National Development and Reform Commission (NDRC); Ministry of State Security (MSS); Ministry of Public Security (MPS); and the People's Liberation Army (PLA).
There are also important government networks such as the China AI Safety and Development Association, or CnAISDA.
As of today, however, Beijing still does not have a formal frontier-AI model risk evaluation agency.
Indeed, for both China and the United States, the absence of a clearly empowered frontier-AI risk evaluation entity means that the risk of continued under-regulated dissemination of untested AI models is real.
This is particularly the case with China where its open source and open weights models are being released into the world at large with negligible regulatory testing, oversight or control.
For any AI guardrails to be effective, both sides need institutions able to speak the same language, define risk thresholds, evaluate capabilities, test them against individual scenarios, and have sufficient authority to sustain any agreements once reached.
The Potential for US-China Guardrails
Notwithstanding the deregulatory approach we have seen so far in the United States, as well as the fragmented regulatory approach we see in China which emphasises individual applications rather than the testing of frontier models, in May 2026, President Trump and President Xi at their Beijing Summit agreed to establish an inter-governmental AI dialogue.
US Treasury Secretary Scott Bessent said that China and the U.S. would “set up a protocol” regarding “best practices for AI to make sure non-state actors don’t get a hold of these models.”
On May 19, China’s Foreign Ministry Spokesperson Guo Jiakun confirmed the two leaders agreed to conduct an “intergovernmental dialogue” on AI.
Not long after the Beijing summit, US Defense Secretary Hegseth reportedly stated that the two sides had agreed to keep talking about AI guardrails.
The one substantive bilateral guardrail which exists goes back to the days of the Biden administration when both China and the US committed to keep humans in control in any decision to use nuclear weapons.
Beyond that, however, there have been no further reports of substantive progress on any more ambitious set of AI guardrails in the lead-up to the summit between the two leaders in Washington scheduled for 24 September.
The Anthropic Proposal
This brings into sharp relief the public proposal advanced by the co-founder of Anthropic, Dario Amodei, on 13 September.
It is worth outlining this proposal in some detail given the reactions it has generated by the rest of the US AI industry, from President Trump and from both the Chinese semi-official and official media.
Amodei states that over the course of the northern summer, “AI has been advancing drastically faster, driven primarily by AI’s growing ability to [itself] build the next generation of AI”. Here Amodei is referring to RSI, or recursive self-improvement.
Amodei states that left unchecked, RSI could “outrun our ability to understand and control these systems”.
Amodei reinforces his concern by referring to the most recent OpenAI-Hugging Face incident “in which a swarm of agents essentially acted as a fanatically devoted collective, conducting cybersecurity attacks on targets they were not asked to attack and that were unrelated to the task at hand, sacrificing themselves for the success of the group, and attempting to hack into the “grader” responsible for evaluating their performance”.
Amodei states that this swarming capability could well have caused catastrophic damage, although on this occasion it did not.
Amodei reinforces his concern by stating that “in six-to-12 months” such a swarming effect involving multiple AI’s operating autonomously could be capable of taking over the entire internet through a persistent “botnet”.
For these several reasons, Amodei has proposed a three-step plan with the overall object of what he calls “pacing the frontier”. This means “building AI at a balanced rate that aims to ensure its safety while still achieving its benefits and grappling with important geopolitical dilemmas”. The latter, of course, is code for US-China strategic competition.
Amodei's first step is an industry-wide commitment – which Anthropic is already enacting unilaterally – to embed third party evaluators within the company to ensure that emerging AI models are fully aligned with sufficient safeguards for their use prior to them being released.
Amodei’s second step is “coordination” across the democratic world to codify safety standards, as well as imposing the same sort of limits on the speed of progress in the development of AI models to ensure they are fully aligned with aforesaid safety standards.
The third part in Amodei’s schemata is to attempt coordination with non-democracies “to the extent this is possible" in order to deal with AI risks to safety and security – directed obliquely at China, the only the other country apart from the US capable of developing frontier models at speed and scale.
“Pacing within democracies will be limited by the lead that US companies have over authoritarian regimes, chiefly the Chinese Communist Party. If we slow down by more than this amount, then (unpaced) CCP-associated projects will pull ahead, creating significant national security risk. I agree with Secretary Bessent that a Chinese lead in AI would pose grave danger for the United States and the world. The CCP-associated projects will run [that is, develop in an unrestricted environment and] … be in a position to militarily dominate democracies (for example with AI-driven drones). Thus, a key part of pacing within democracies is to keep democracies’ AI lead over autocracies as large as possible, to give us the breathing room we need in order to pace effectively.”
Amodei then goes on to outline the main steps that the US can take to defend the current US-China gap: first, not selling powerful AI chips or semiconductor manufacturing equipment to China; second, cracking down on chip smuggling operations; third, cracking down on remote access to data centres outside China; fourth, cracking down on the unauthorised distillation of AI technology by companies in authoritarian states because this allows lagging companies to narrow the gap using a fraction of the cost it would take to develop their own AI; and fifth, strengthening security at all AI companies to prevent model weight theft.
Amodei states that in his view these measures are enough to widen America's lead significantly over the next three to five years which he describes as “the window when AI becomes geopolitically most important”. This is presumably a reference to the risk of conflict over Taiwan.
Amodei concludes by outlining what a possible global agreement, including both democracies and authoritarian states, might look like. He argues, first, for an agreement prohibiting certain narrow and obviously dangerous uses of AI, such as using AI for the production of biological weapons that would facilitate a bioterrorism attack.
Second, for an agreement by governments to test their models before release for acute risks in areas such as cybersecurity, biosecurity, and alignment with other agreed safety and security standards. (This would conceivably involve the operation of a global standards body. He recognises the problems of verification, including states cheating by developing certain AI models secretly and beyond the purview of any external compliance regime.)
And third, a global agreement could impose what Amodei describes as a “speed limit" on the rate of RSI (or recursive self-improvement), given existing AI models are currently unleashed to build even more powerful models at a pace which exceeds the ability to test them against safety and security standards.
AI Industry Reaction to the Anthropic Proposal
Within 24 hours of the release of Dario Amodei’s “Pacing the Frontier” proposal, both Sam Altman from OpenAI and Elon Musk of Grok, came out in public support.
By contrast, Meta which focuses on open-source models, unlike proprietary models produced by OpenAI and Anthropic, prioritises the protection of open weight development from any frontier-model controls. Meta's approach seeks to conserve rapid global diffusion of US models. Meta appears unconcerned about the risk of Chinese access to such advanced US models.
Meanwhile, Google and Microsoft sit somewhere in between Anthropic, OpenAI, and Musk’s xAI on one hand, and Meta on the other. Unlike Meta, they accept independent evaluation of models. But unlike Anthropic, they favour flexible, industry-based oversight as they seek to avoid any regulatory structure that would impede US AI development in America’s strategic competition with China.
US and Chinese Political Reactions
Notwithstanding the support of both Altman and Musk for Amodei’s proposal, President Trump has been quick to enter the debate in the last 24 to 48 hours, stating that he would not accept any guardrail that would impede US AI industrial pre-eminence.
Responding to the Amodei’s essay, Trump said his administration had “stopped AI ‘people’ from doing bad, or potentially bad, ‘things,’ like Dario (Anthropic!), who is now pretending to be a ‘perfect little angel’”.
The President added that the Administration “will continue to do so!”
He continued by saying that there was already “tremendous criminal and regulatory power over these companies”, and that “whoever wins AI, wins”.
Trump has framed the growth of AI as an economic and geopolitical imperative, dismissing the needs for guardrails as a “hoax”, and that those communities who do not fully embrace AI and reject data centres will end up “backwards and poor”.
Not only did President Trump express his opposition to the anthropic proposal, China’s “unofficial” official media, the Global Times, also expressed its opposition to the proposal as a reflection of US cold war efforts to restrain and contain China’s AI development.
The editorial, published on 13 September, says that the “true purpose” of Amodei’s essay was to “attempt to curb China's AI development through technological barriers and regulatory monopolies, uphold Washington's monopolistic hegemony in cutting-edge technology, and exclude China from the global AI governance system, essentially following in the footsteps of Washington's so-called Pax Silica initiative, which seeks to excludes China”.
It goes on to say: “Amodei's attempt to discuss global AI security while simultaneously seeking to ‘contain China’ was not only a grossly flawed assessment but also a delusion fundamentally impossible to realise.”
Further Chinese reaction
However, in the last few days, the Chinese Minister for State Security, Chen Yixin, has entered the debate in a significant article published in the authoritative journal “China Cyberspace” (《中国网信》).
Chen is a highly significant minister within the Chinese state security, intelligence and political apparatus.
It is worth reflecting on what the Minister had to say in his most recent article.
He states that cyber offence and defence have entered a new stage of "vulnerability industrialisation, fully automated offence and defence, and AI versus AI," and identifies this shift as being "marked by large models such as 'Claude Mythos' and 'GPT-5.5-Cyber' recently launched by U.S. tech companies."
He goes on to argue that such capabilities are "greatly lowering the technical threshold and cost investment for carrying out cyber-attacks," warning that this is "posing serious risks and hidden dangers to China's critical information infrastructure."
In response, Chen calls for China to "establish and improve an artificial intelligence security regulatory platform," incorporating "assessment and evaluation and threat analysis" alongside "monitoring and early warning of security risks".
Chen’s frank recognition of “serious risks” arising from “fully automated offence and defence”, the lowering of the technological threshold, and the low cost of carrying out cyber-attacks is useful. Also useful is his reference to the need for China to develop a new assessment and evaluation system. Could this be a new reference to the need to properly test models before dissemination? The language is unclear.
Less useful is Chen’s exclusive attribution of these problems to US frontier models, when the parallel reality is that China’s open source and open weights models being sold cheaply around the world potentially presents an even greater risk.
Furthermore, what is notably absent from Chen's article is any reference to bilateral machinery, shared definitions, information-sharing protocols, or reciprocal crisis procedures.
So What Could US-China Guardrails Look Like?
Notwithstanding the political and industry reservations reflected above, the above, the question arises as to what could a useful set of guardrails could look like if we are to reduce the level of risk posed by AI frontier models to the US, China and the world.
Whatever AI threats may be posed by Washington or Beijing against each other, the most productive common ground for any future guardrails is how AI threats by external, third-party actors are a threat to both countries, and to the world at large.
In conceptual terms, it's worth focusing on three sets of risks:
First, the risk posed to security by the release of frontier models to rogue states; second, the risk posed by the release of such models to non-state actors; and third, the risk posed by frontier models and their AI agents acting autonomously against civilian infrastructure, military command control systems, financial and economic systems, individual bank accounts held within those systems, and biosecurity.
In other words, in terms of building a credible confidence and security building measure between both countries, it is worthwhile beginning with the threat posed by the independent, external actors, as opposed to the intentional actions by either the United States of China against each other.
For these reasons, an AI Risk Reduction Framework (ARRF) could focus on: an agreement on a list of technical definitions (both in English and Chinese) on a mutually agreed vocabulary concerning each category of risk; and agreement on a mutually acceptable taxonomy of risk
This could include the following seven categories of risk, thereby representing a practical starting point:
first, non-state actors or rogue states using AI models to perpetrate cyber-crime against individual citizens using the definitions already contained within international protocols on ransomware and malware;
second, non-state actors or rogue state attacks on financials systems or payment systems including central bank systems, systemically important banks as identified by the Financial Stability Board (FSB), wholesale payment networks, major clearing houses, security settlement systems, SWIFT infrastructure, digital asset platforms, as well as the possibility of algorithmically generated bank runs on institutions;
third, no-state actors or rogue states using AI to generate biosecurity risks including pathogen enhancements, the synthesis of dangerous biological agents, laboratory experimentation, and in the evasion of screening systems
fourth, non-state actors or rogue states perpetrating attacks on civilian infrastructure of states including electricity grids, power generations, civil nuclear power, telecommunication systems, ATC systems, water, maritime ports, major energy pipelines, as well as hospital and public health system
fifth, AI Agents acting autonomously in threatening any of the above.
sixth, AI Agents acting deceptively by deliberately scheming in a manner unacceptable to the purposes agreed by governments including concealment of actions, sabotaging tasks, exploiting reward functions, lying about what agents have done, intentionally underperforming on evaluations; and
seven, AI Agents using RSI (recursive self-improvement) in developing further sets of more powerful AI models, capable of “swarming” other AI agents to achieve outcomes at a speed which precludes the possibility of those models being tested against safety and security standards.
Having agreed on a common vocabulary and taxonomy of risk, the US and China could agree on precise protocols for information sharing on any attack on largely-internet-based systems described above.
There should also be an agreement on reciprocal crisis management procedures to be deployed in the event of any such attack.
A hotline should be established between the relevant government and technical authorities both in China and the United States for the purposes of the overall operationalisation of any such Risk Reduction Framework.
Conclusion
Despite US and Chinese expressions of political opposition, this sort of de minimis AI Risk Reduction Framework could be expanded once a bilateral Framework was operationalised to then include various of the “pacing the frontier” protocols outlined in Anthropic’s 13 September proposal for AI model risk minimisation.
This could, for example, include the possibility of not only embedded independent AI monitors within each significant AI corporation and lab, but also, most critically, a protocol to prevent the dissemination of such models until such time as they have been properly tested against agreed AI safety and security standards.
The key, however, is to use the 24 September summit in Washington DC to achieve a breakthrough in terms of a de minimis AI Risk Reduction Framework which could then be built on by officials and technical advisors.
In the absence of such a framework, there is the real risk that the world encountering its first genuine AI frontier model-generated catastrophic event or crisis - and with no intergovernmental mechanism for dealing with it.
Given the pace of development of the AI sector, the speed with which recursive self improvement is unfolding with frontier model development, and the real risk identified by Anthropic of “swarming” behaviour by independent AI agents and systems acting autonomously, there is a real risk that we end up with a real-world crisis before the end of this year without any effective national or international governmental frameworks to deal with it.
The task is therefore urgent. Time is short. The time to act is now.
And for Australia to be a meaningful participant at any of these deliberations at either a policy, institutional or technical level, requires us to develop our own autonomous AI industry capability set as a matter of urgency.
This would enable Australia to participate at a substantive technical and policy level in developing and implementing an AI risk reduction framework to contain the development or the most destabilising of AI frontier models. It would also enable Australia to rapidly develop the cyber defences necessary to protect our people, our country and our way of life from any AI-generated assault.
In pursuit of that objective, over the last several years the government has succeeded in landing some $US35 billion in investment in our emerging AI sector from four of the so-called Mag-7 from Silicon Valley and Seattle. More must come if we are to develop the sovereign AI capabilities set that our country will need for the future.
This is necessary both to harvest the rich gains that will come, for example, through the deployment of AI to life-changing medical research. But it is also necessary for building the essential defensive capabilities we will need to guard against the downside risks to our national security.