Submission 496
Mr Matthew Vitale
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.
The author, an AI software company co-founder writing personally, argues for measurable Commonwealth AI spending, adoption costs and domestic capability obligations. Identifies price as a barrier but says published data do not test this proposition. Asks for space for open-weight providers, conditional on measured outcomes, and a consumer AI data-rights assessment, explicitly allowing that designation may be wrong.
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.
The author argues that Commonwealth AI spending, and the share reaching Australian-owned suppliers, cannot be determined from published data.
How much does the Commonwealth spend on artificial intelligence, and how much of that reaches Australian-owned suppliers. Neither can be answered from what the Commonwealth publishes.
The author argues that the published measure of Australian participation in procurement records supplier address rather than ownership.
A wholly foreign-owned supplier with a Sydney office has an Australian address. The Government's own measure of Australian participation in its own buying cannot distinguish an Australian-owned business from an Australian letterbox.
Says adoption patterns are consistent with cost being a material barrier, but published data do not test that proposition.
That pattern is consistent with cost being a material barrier, but the published data does not test the proposition.
The author argues that a sovereignty purpose has been adopted, with obligations enumerated for energy, water, land use and skills but none for compute access or domestic capability.
A purpose has been adopted, and obligations are being enumerated for energy, water, land use and skills. None has been enumerated for access to compute or for domestic capability, and nothing measures delivery of either.
Asks government to leave room for providers like the author’s company, then measure customer control, manageable costs and retained domestic capability; says providers failing that test should not benefit.
I am not asking the Commonwealth to prefer providers like mine. I am asking it to leave room for them and then measure whether they produce anything: control a customer can exercise, costs a customer can manage, and capability that stays in the country. A provider that cannot meet that test should not benefit from it, and that includes mine.
Argues consumer AI records create switching costs and lack machine-readable portability rights, with Part IVD already providing an assessment process.
On paragraph 1(i), the record a consumer AI assistant accumulates creates real switching costs, and Australian law gives a person no machine-readable right to move it. Part IVD of the Competition and Consumer Act already provides the process for deciding whether it should.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Initiate a Part IVD assessment of consumer AI assistants: record-class specification, machine-readable access, enforceability given record location, and security/privacy/cost consequences, including whether inter-provider transfer should be permitted at all.
recommend that the Commonwealth initiate the sectoral assessment for which Part IVD of the Competition and Consumer Act already provides, directed at consumer AI 4 assistants, and that its terms require the assessment to reach four questions: whether the record a provider accumulates about a person is capable of being specified as a class of information; whether a person should be able to obtain it in a form a machine can read; whether the place the record is held leaves any resulting right enforceable; and what security, privacy and cost consequences would follow, including whether transfer between providers should be permitted at all;
Make Commonwealth AI buying measurable by recording on every contract notice whether the procurement involves AI, the expected value estimated when selecting the procurement method, and whether the successful supplier meets the Commonwealth Procurement Rules definition of an Australian business.
recommend that the Commonwealth make its own AI buying measurable, by recording on every contract notice whether the procurement involves artificial intelligence, the expected value the official estimated when the procurement method was chosen, and whether the successful supplier meets the Commonwealth Procurement Rules definition of an Australian business;
Publish each year, in existing procurement statistics, the number and value of AI contracts awarded to Australian businesses, the counts and values within the bands the rules turn on, and how often the Australian business rule was set aside.
by publishing each year, in the procurement statistics that already exist, the number and value of AI contracts awarded to Australian businesses, the counts and values falling within the bands the rules themselves turn on, and how often the Australian business rule was set aside;
Annually measure and publish how many Australian businesses adopt AI and on what terms, including adoption and retention costs for firms employing 20–199 people.
recommend that the Commonwealth measure and publish annually not only how many Australian businesses have adopted AI but on what terms, including what it costs a business employing between 20 and 199 people to adopt and to retain AI capability, so that who is being priced out is known rather than assumed;
In the Australian Standards for AI, write any expectation about local capability as an obligation that can be measured and reported publicly rather than as a statement of intent.
recommend that, when the Australian Standards for AI are drafted, any expectation about local capability be written as an obligation that can be measured and reported publicly rather than as a statement of intent,
Name established mid-sized firms, including those employing 20 to 199 people, among the businesses a local-capability expectation is meant to benefit, not only start-ups, small businesses, researchers and not-for-profits.
and that the businesses it is meant to benefit be named to include established mid-sized firms, including those employing between 20 and 199 people, and not only start-ups, innovative small businesses, researchers and not-for-profits;
Measure whether alternative providers deliver customer control, manageable costs and retained Australian capability; withhold benefits from providers failing that test, including the author’s company.
I am not asking the Commonwealth to prefer providers like mine. I am asking it to leave room for them and then measure whether they produce anything: control a customer can exercise, costs a customer can manage, and capability that stays in the country. A provider that cannot meet that test should not benefit from it, and that includes mine.
In setting AI standards and buying AI, apply and publish a test of whether the requirement can be met by an Australian provider running an open-weight model in the customer's own environment.
recommend that, in setting standards for artificial intelligence and in buying it, the Commonwealth apply and publish one test: whether the requirement it is imposing can be met by an Australian provider running an open-weight model in the customer's own environment.