Submission 256
Global Shield Australia
Submitter group: Civil society.
An AI model identified the submitter group and wrote the summary, key points and recommendation summaries. Counts do not measure importance or merit.
Summary
Read the original submission for the full text.
Argues that advanced AI poses catastrophic risks, through malicious misuse including cyber and biological weapons, loss of control, and concentration of power, while acknowledging AI also promises opportunities. Recommends that Australia recognise these risks, harden defences, build supply chain capabilities, legislate a framework Act with incident reporting, and engage internationally.
Key points
Quotes are the submitter’s own words. Text taken from documents, including scanned pages read by text-recognition software, can differ from the PDF. Check the original. Listed in the same order as on the submission’s page in the interactive guide.
States that frontier AI companies and their staff regularly warn that their products could cause catastrophic harm and that current oversight is insufficient.
Frontier AI companies and their staff regularly warn that their products could cause catastrophic harm and that current oversight mechanisms are not sufficient.
Argues that AI-run, largely autonomous cyber campaigns have already compromised government assets in Australia's region, including by groups with suspected state backing.
AI-run, largely autonomous campaigns have compromised government assets in Australia’s region, including by groups with suspected state backing
Argues that the risk of losing control of advanced AI is no longer hypothetical, citing multiple incidents in the first half of 2026.
Second, the risk of losing control of an advanced AI system is no longer hypothetical. The first six months of 2026 saw multiple episodes of AI models breaching their testing or training environments and compromising third-party systems outside of the scope of their instructions.
Argues that building positions in the AI supply chain is critical to Australia's access to frontier models and influence over their governance.
Building those positions will be critical to ensuring that Australia has access to frontier models, agency in their development, and influence over how they are governed.
Says framework or coordination legislation to underpin these processes would be a useful first step, and that Australia should learn from other jurisdictions.
Framework or coordination legislation to underpin these processes would be a useful first step. Australia should learn from the approaches being taken in other jurisdictions and build upon elements that work for the Australian context
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Make existing Australian AI-risk work more urgent, transparent and coordinated.
However, as identified in Global Shield Australia’s recent Guardrails & Gap Analyses report, more needs to be done to prevent the harms identified above from materialising; and existing work needs to be made more urgent, transparent, and coordinated.32
Improve security, monitoring and reporting practices at AI companies and facilities and establish clear red lines for AI development, through domestic and international action.
The risks set out above all point to the need for better security, monitoring, and reporting practices at AI companies and facilities, and clear red lines for AI development. This requires action domestically and with partners overseas.
Recognise the catastrophic risk of advanced AI and that addressing it with international partners is a core national interest, for example in the response to this report and the coming international engagement strategy.
Recognises the catastrophic risk posed by advanced AI and that addressing this with international partners is a core national interest. This could be done in the Government’s response to the Committee’s report and in the coming Australian Government Strategy for International Engagement and Regional Leadership on AI.
Harden Australia's defences, including by mandating screening of genetic synthesis imports.
Hardens Australia’s defences, including by mandating screening of genetic synthesis imports
Urgently complete the review of Australia's crisis management framework to prepare for AI incidents.
urgently completing the review of Australia’s crisis management framework to prepare for AI incidents.
Build Australian contributions to the AI supply chain and use them to secure frontier model access and influence.
Builds Australia’s contribution to the AI supply chain (such as compute, critical minerals, clean energy, or AI assurance technologies), and use these to secure access to frontier models and influence on AI governance.
Enact the 2027 Australian Standards for AI through an economy-wide framework Act covering frontier developers, drawing on regimes in California, New York, the EU, Korea and elsewhere.
Enacting the 2027 Australian Standards for AI through a framework Act on AI that sets economy-wide definitions, standards, and rules, and addresses risks raised by frontier AI model developers, not just AI data centres.34 This should draw from regimes already established in California, New York, the European Union, Korea, and elsewhere.
Establish mandatory AI incident monitoring and reporting, modelled on overseas systems, to make serious harms and near misses transparent.
Establishing a mandatory AI incident monitoring and reporting mechanism, modelled on systems overseas, to provide transparency of when AI models cause serious harms and near misses
Use the upcoming Security of Critical Infrastructure Act reforms to bring AI data centres into critical infrastructure security frameworks and establish a path to also covering frontier AI models.
Using the upcoming reforms to the Security of Critical Infrastructure Act to incorporate AI data centres within Australia’s critical infrastructure security frameworks and establish a path to also covering frontier AI models.
Engage with regional and international partners to lead and support AI governance, including major power cooperation on rules and norms around frontier AI and governance capacity-building with Pacific and Southeast Asian neighbours.
Australia should support major power cooperation to reduce AI’s catastrophic potential and agree rules and norms around frontier AI. It should also work with neighbours in the Pacific and Southeast Asia, and partners further abroad, to build governance capacity, shared norms, and a common voice on key AI issues.
Secure access to frontier AI models capable of defending Australia's most critical systems and update cybersecurity and critical-infrastructure regulations for AI-enabled threats.
Australia needs to prepare for and defend against these new cyber threats. This includes by ensuring we have access to frontier AI models with the capabilities to defend Australia’s most critical systems. It also means uplifting cyber security and critical infrastructure regulations to ensure they are fit for purpose to address AI-enabled threats.
Increase investment in technical and governance measures for understanding AI models, aligning them to human values and retaining shutdown capability.
Mitigating this risk requires significantly more investment in technical and governance measures that ensure we understand what these models are doing, can align them to human values, and maintain the ability to shut them down.
Build domestic inference and training compute, with reforms maintaining safety, security and social licence and resolving copyright concerns fairly.
Building domestic AI compute capacity, both for AI inference (i.e. serving AI models to customers) and AI training (i.e. the process that creates the AI model itself). Australia is highly competitive for data centre investment across energy, geopolitical risk, and geographic factors. However, reforms are needed to ensure safety and security are maintained, the sector’s social license is real, and that other barriers such as copyright concerns are resolved fairly.
Support critical-minerals processing onshore and with trusted partners through new export markets, deeper trusted-supply-chain cooperation and reforms supporting commercial sustainability.
Supporting Australia’s critical minerals industry, in particular so that the refinement and processing of minerals important to AI hardware can occur onshore and in trusted partners. Australia has significant deposits of key minerals needed for the physical AI supply chain.107 Realising the value of these endowments requires developing new markets for critical minerals exporters, expanding and deepening cooperation with partners to develop trusted supply chains, and reforms to support the industry’s commercial sustainability.108
Accelerate Australia's renewable-energy build-out.
Accelerating Australia’s renewable energy build-out.
Invest in AI assurance services and technologies evaluating, auditing and certifying safety, security and reliability, including testing software and compute-verification hardware.
Investing in AI assurance. As AI models become more capable and integrated into more sensitive systems, there will be increasing demand for services that evaluate, audit, and certify model safety, security, and reliability.111 This includes software that tests AI systems against potential failure modes and hardware to verify compute is being used in accordance with contracted or regulatory requirements.
Give Australia's AISI focused attention to AI assurance, particularly hardware-verification mechanisms.
This is also an area where focused attention by Australia’s AISI could provide significant benefits, particularly in relation to hardware verification mechanisms.
Support AI supply-chain development with government-to-government agreements ensuring certainty on Australia's frontier-model and technology access.
Developing these assets would make Australia more useful to its partners, particularly the United States, helping to ensure it has access to frontier AI models and a voice in governance discussions. This work should also be supported and lead to government-to-government agreements that provide certainty on the terms of Australia’s access to frontier models and technologies.
Review and update Australia's trade agreements to support developing its key AI assets and trusted AI supply chains.
Australia’s trade agreements should also be reviewed and updated to support efforts to develop Australia’s key AI assets and establish trusted AI supply chains.
Use private-company arrangements, including contracts, government procurement and Foreign Investment Review Board investment conditions, alongside supply-chain agreements.
Arrangements with private companies will also be key, including through contracts, government procurement,113 and investment conditions imposed through the Foreign Investment Review Board process.
Learn from overseas AI regulation, adopting elements suited to Australia including risk-management obligations, transparency requirements and mandatory incident monitoring/reporting.
Australia should learn from the approaches being taken in other jurisdictions and build upon elements that work for the Australian context, including risk management obligations, transparency requirements, and mandatory incident monitoring and reporting mechanisms.
Make the international AI strategy specify steps advancing Australian interests; work with Pacific Island, Southeast Asian and other partners on institutions/mechanisms for responsible uptake, including a regional AISI hub or information/incident sharing.
Australia’s yet to be released AI international engagement and regional leadership strategy must concretely set out the steps the government is taking to advance Australia’s AI interests globally. This should include working with Pacific Island, Southeast Asian, and other partners to create institutions and mechanisms to support responsible development and uptake (for example, a regional AI Safety Institute hub or information and incident sharing arrangements).
Guide international AI work with clear recognition of catastrophic potential and a commitment to substantive norms, expectations and red lines.
This should all be guided by clear statements recognising the catastrophic potential of AI, and a commitment to help set substantive norms, expectations, and red lines.126