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 548

Monash eResearch Centre

Submitter group: Universities.

An AI model identified the submitter group and wrote the summary, key points and recommendation summaries. Counts do not measure importance or merit.

Open in the interactive guide

Summary

Read the original submission for the full text.

A Monash eResearch director describes how the centre uses AI in research and its own software and data work, and argues for secure national access to commercial AI services alongside funded onshore sovereign open-weight capability for sensitive data, citing Monash's MAVERIC supercomputer as an example. States this is not a rejection of commercial models. Makes four recommendations for consideration.

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.

  1. Says most researchers need safe, reliable access to capable models, compute and expert help for their data, rather than building new models.

    Most researchers do not need to build new AI models; they need reliable, safe access to capable models, with compute and expert help to apply them to their own data.

  2. Describes using AI on the principle that a named person remains accountable for every output, with staff more productive but their time shifting to review.

    We use AI on one principle: a named person remains accountable for every output, irrespective of AI use. This does mean our staff are more productive, but their time shifts to performing more of a review role.

  3. Says AI tools that can find vulnerabilities could be misused, and that its engineers use the same tools defensively to review code, check configurations and monitor systems.

    AI tools that can find software vulnerabilities could be misused against systems like ours, but our engineers can use the same tools defensively. AI now helps them review code for security flaws, check configurations and monitor systems

  4. Says Monash's MAVERIC can support data control, reproducible model versions, fine-tuning under Australian governance and protection from foreign providers changing terms, pricing or availability.

    This is a good example where Monash's AI supercomputer, MAVERIC, can play a very important role. Data never leaves our control; a model version can be fixed and rerun years later for reproducibility; models can be fine-tuned on Australian data under Australian governance; and research is not exposed to a foreign provider changing its terms, pricing or availability.

  5. Describes the approach as balanced between commercial services and sovereign open-weight capability, with compute capacity as the constraint.

    This is a balanced approach, not a rejection of commercial models: commercial services where data and risk allow, sovereign open-weight capability where they do not. The constraint is capacity.

Recommendations

Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.

  1. Negotiate secure national commercial AI access, such as Australian hosting and no training use, for government, health services, research, industry and smaller organisations where data and risk allow.

    Negotiate national access to commercial AI services on secure terms, such as data hosted in Australia and not used for model training, so government, health services, research, industry and smaller organisations can use leading models where data and risk allow.

  2. Fund shared onshore sovereign compute, open-weight models and engineers to run and secure them, for sensitive data and critical functions where commercial services are inappropriate.

    Fund shared, onshore sovereign AI capability, covering compute, open-weight models and the engineers to run and secure them, for sensitive data and critical functions where commercial services are not appropriate.

  3. Issue practical national guidance on human oversight of AI covering proportionate review, disclosure and record keeping.

    Issue practical national guidance on human oversight of AI: proportionate review, disclosure and record keeping.

  4. Help software-writing or sensitive-data organisations, including small businesses, health services and local government, adopt defensive AI through shared tools, threat intelligence and practical guidance.

    Help organisations that write software or hold sensitive data, including small businesses, health services and local government, to adopt AI defensively through shared tools, threat intelligence and practical guidance.

Read the original

Open in the interactive guide