AI Inquiry Atlas

An unofficial, independent guide to the public records of the Joint Select Committee on Artificial Intelligence. Not connected with the Parliament of Australia or the committee. Its makers also made a submission (no. 481; see About). Summaries and scores are written by AI and can be wrong: check the original before relying on them.

Submission 443

Name withheld

Submitter group: Individuals.

Position labels are not shown for individual submitters.

An AI model identified the submitter group and chose the extracts. Counts do not measure importance or merit.

Open in the interactive guide

Extracts from the submission, in the submitter’s own words.

Extracts

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.

  1. Cutting graduate intake is correct for almost any single firm looking at its own next three years. It is catastrophic when every firm does it at once.

  2. Residency is not sovereignty. Storing data in an Australian region of a foreign-owned platform tells you where it sits, not whose law governs access to it. A provider domiciled elsewhere remains subject to that country's compulsion powers regardless of server location.

  3. Australia will not win a competition to build frontier general-purpose models, and pursuing it would spend public money at the layer commoditising fastest. Australia can win on narrow, domain-specific models trained on Australian data in areas of real sectoral expertise.

  4. the cost of prevention has been transferred to victims, parents, schools and small businesses who cannot bear it. This will not be corrected by better consumer education. It requires obligations on the parties actually positioned to prevent the harm.

  5. The problem is enforcement, not permission. Australia has no general text and data mining exception of the kind adopted in the European Union, Japan and Singapore, and Australian fair dealing is purpose- limited and does not obviously extend to commercial model training.

  6. I am not asking the Committee to adopt a position on whether these estimates are correct, and I would be sceptical of any submission claiming certainty in either direction. I am asking that Australia hold the capability to form its own view.

Extracts from recommendations

Extracts chosen by an AI model, in the submitter’s own words. Listed in the same order as on its page in the interactive guide.

  1. The framing I would ask the Committee to hold throughout is this: the interests of model owners and the interests of the Australian public are not the same interests.

  2. Where those interests overlap, adoption should proceed and government should enable it.

  3. I would also ask the Committee to reject the framing that caution costs competitiveness.

  4. Of the twenty-three recommendations at the appendix, I would ask the Committee to give particular weight to six: 1, 7, 13, 17, 19 and 21.

  5. I would ask the Committee to separate speculative future risk from present, measured, compounding harm. These harms are occurring now, at scale, to identifiable Australians.

  6. The response must move from individual detection, which is no longer possible, to verification infrastructure and liability. Where a payment provider, carrier or platform is the channel through which a loss occurs and is positioned to interrupt it, the cost of failure should sit with the operator rather than the victim.

  7. These are subjective judgements rather than measurements, and I would ask the Committee not to treat them as data.

  8. That the ABS, with Treasury and the Productivity Commission, establish and publish a national measurement framework for the labour market effects of AI, disaggregated by age cohort, entry-level occupation, industry and region, reported annually.

  9. That Commonwealth tax concessions, grants and procurement preferences relating to AI be conditional on the recipient maintaining or increasing graduate, apprentice and entry-level intake.

  10. That a defined trigger for intervention be established, so targeted measures follow automatically if graduate hiring in AI-exposed occupations falls beyond a specified threshold.

  11. That publicly funded AI skills programs be directed at evaluation and verification capability rather than tool proficiency.

  12. That Commonwealth agencies deploying AI in professional functions maintain, exercise and report on unaided capability in those functions.

  13. That a statutory distinction be legislated between data residency and data jurisdiction, requiring vendors to disclose every jurisdiction under which source data and derived artefacts may be compelled, and that procurement assess jurisdiction rather than server location.

  14. That a classification framework identify categories of Australian data, including health, critical infrastructure operational data, Indigenous cultural material and government policy material, that may not be processed outside Australian jurisdiction without explicit approval.

  15. That concentration of dependency on a small number of AI providers be monitored as a systemic risk, on a basis comparable to financial system supervision.

  16. That Commonwealth agencies and critical infrastructure operators maintain an auditable AI data flow register.

  17. That a national strategy treat high-value Australian domain-specific datasets as strategic assets to be protected, licensed and commercialised.

  18. That Commonwealth scam awareness messaging be rewritten, on the basis that advice to identify fraud through language quality is now obsolete and counterproductive.

  19. That Australia adopt a mandatory reimbursement regime for authorised push payment fraud, allocating liability according to capacity to prevent the loss.

  20. That liability for non-consensual synthetic sexual imagery extend beyond the individual creator to the providers, app stores, payment processors and hosting services that make such applications available to Australians.

  21. That eSafety and Commonwealth law enforcement receive funding scaled to the volume growth of synthetic child sexual abuse material, with triage and victim identification capability reported to Parliament annually.

  22. That age assurance and minimum age protections extend to conversational and companion AI, with a statutory duty of care on providers rather than on children and parents.

  23. That a framework be developed for provenance, authentication and evidentiary treatment of synthetic media, in consultation with the courts, law enforcement and the media sector.

  24. That disclosure of training data provenance be mandated for any AI model offered commercially in Australia, at a granularity sufficient for a rights holder to determine whether their work was used, with enforceable penalties.

  25. That Australia decline to introduce a broad text and data mining exception, and affirm that commercial model training is a use requiring licence.

  26. That a statutory licensing scheme be established for the use of Australian creative, cultural and media content in AI training, administered through existing collecting society infrastructure, with remuneration flowing to creators.

  27. That a bargaining framework be considered for the use of Australian news and journalistic content by AI systems.

  28. That Indigenous cultural and intellectual property receive specific statutory protection in the context of AI training and generation, developed with and controlled by Aboriginal and Torres Strait Islander communities.

  29. That the Australian AI Safety Institute be placed on a statutory footing with powers to compel information from developers of frontier systems offered in Australia, including pre-deployment evaluation results, and be resourced to conduct independent assessment.

  30. That a standing national AI risk assessment be established and reported to Parliament annually, covering capability change, concentration, critical infrastructure dependency and catastrophic risk, so that Australia holds an independent view rather than inheriting one.

Read the original

Open in the interactive guide