Submission 191
Mr Eli Hayes and Dr Darren Lim
Submitter group: Individuals.
Position labels are not shown for individual submitters.
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 AI sovereignty means maximising Australia's decision space rather than self-sufficiency. Distinguishes coercion from access restriction, assesses domestic compute's resilience benefits and limited leverage, and examines Pax Silica without finding enough concrete Australian outcomes to assess it. Recommends five tests for sovereignty proposals, including reciprocal access and benefits proportionate to exclusive commitments.
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.
Argues that self-sufficiency in AI is neither plausible nor necessary; the goal should be retaining sovereignty, defined as maximising Australia's decision space.
Self-sufficiency in AI is neither plausible nor necessary – Australia will remain highly dependent on other countries and firms, even as the scale and nature of this dependency will change in different parts of the stack. Instead, the goal should be retaining sovereignty, which we define as maximising the decision space available to Australia.
Describes restriction as a state unilaterally revoking access to part of the AI stack for reasons unrelated to its relationship with Australia.
By contrast, restriction is when a state unilaterally revokes access to part of the AI stack for reasons that are unrelated to their relationship with Australia.
Argues that local compute is necessary but not sufficient for a resilience strategy if access to foreign-hosted models is disrupted.
While local compute is not sufficient by itself, it is a necessary component of any resilience strategy that seeks to preserve AI functions if access to foreign-hosted models is disrupted.
States that resilience is the first line of defence against coercion and restriction, but cannot remove dependence on foreign AI capabilities altogether.
Resilience is the first line of defence against both coercion and restriction. But resilience cannot remove Australia’s dependence on foreign AI capabilities altogether, nor would complete technological self-sufficiency be desirable or realistic.
States that Pax Silica has not yet produced sufficiently concrete Australian-specific outcomes to assess it against the submission's criteria.
Pax Silica has not yet generated sufficiently concrete Australian-specific outcomes to assess it against these criteria.
Argues that domestic hosting should not be equated with strategic leverage over other actors.
Australian-based compute may improve resilience and strategic positioning without giving Australia meaningful coercive power.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Maintain a strong distinction between coercion and restriction in AI dependency, with differentiated approaches to each risk.
With this in mind, we urge the following: I. A strong distinction between coercion and restriction and differentiated approaches for both risks.
Require decisions that restrict sovereignty to demonstrate greater benefits to balance against the risks.
Given the emphasis on retaining sovereignty, decisions that restrict sovereignty should be required to demonstrate greater benefits to balance against the risks.
Committee to consider, when assessing sovereignty proposals, identifying the policy objective: resilience, strategic positioning or retaliatory leverage.
Identify the policy objective: What problem is a solution trying to solve? Is it about resilience, strategic positioning, or retaliatory leverage?
Committee to consider testing the reliance case against the consequences of disruption: onshoring or duplicating capability is most valuable where loss of access would be highly damaging and alternatives costly or slow.
Test the reliance case against the consequences of disruption: Onshoring or duplicating capability is most valuable where loss of access would be highly damaging, and alternatives would be costly or slow to obtain.
Assess domestic compute for resilience and strategic positioning without equating hosting with leverage; leverage requires disruption to impose significant costs on the actor Australia seeks to influence.
Do not equate domestic hosting with strategic leverage: Australian-based compute may improve resilience and strategic positioning without giving Australia meaningful coercive power. It creates leverage only where disruption would impose significant costs on the actor whose behaviour Australia seeks to influence.
Assess international arrangements by their contribution to Australian decision space, testing reciprocal access, predictability, collective bargaining capacity and preservation of alternatives.
Assess international arrangements by what they add to Australian decision space: Relevant tests include reciprocal access, predictability, collective bargaining capacity and preservation of alternatives.
Apply higher evidentiary thresholds as international commitments become more exclusive; justify deeper exclusivity only where credible benefits scale to the commitment.
Apply a higher evidentiary threshold as commitments become more exclusive: Arrangements with deeper exclusivity can be justified, but only if the credible benefits scale to the commitment.