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 319

Google

Submitter group: Companies.

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.

Google Australia outlines its $1 billion Digital Future Initiative covering research, infrastructure, healthcare, climate and workforce applications of AI. It argues that existing, largely technology-neutral statutory regimes already govern AI deployment across sectors, and recommends sector-based oversight with a central technical hub, high-risk guardrails limited to high-consequence uses, and alignment with international standards.

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. Google describes its $1 billion, five-year Digital Future Initiative as an investment in Australia's sovereign digital foundation.

    The Digital Future Initiative (DFI) represents a landmark $1 billion, five-year strategic investment in Australia’s sovereign digital foundation.

  2. Describes AI research into diabetic-retinopathy screening in remote regions, particularly for Indigenous people in Western Australia.

    Google has also partnered with the Lions Eye Institute to research AI solutions for detecting diabetic retinopathy. This partnership is particularly focused on the Indigenous population in Western Australia and aims to deliver rapid, point-of-care screening results in remote regions to enable early intervention and prevent avoidable blindness.

  3. Describes support for Australian emergency services and researchers to prepare for FireSat bushfire data.

    In Australia, AFAC (the Australian and New Zealand Council for fire and emergency services) and CSIRO are collaborating with Earth Fire Alliance, and with support from Google.org and Google Australia, to help the nation prepare for FireSat data.

  4. Google argues that prescriptive mandates and regulatory ambiguity delay AI adoption and burden startups and SMEs, and that high-risk guardrails should be tailored to high-consequence uses.

    Prescriptive mandates and regulatory ambiguity delay AI adoption and disproportionately burden local startups and SMEs. As highlighted in international analyses like the Draghi Report, heavy compliance demands create barriers that entrench incumbents. High-risk guardrails should be strictly tailored to high-consequence use cases, leaving everyday business applications unencumbered.

  5. Google argues that general purpose AI models are not inherently high-risk and that static compute metrics are poor indicators of capability.

    General Purpose AI models are not inherently high-risk; their risk profile depends on real-world application. Additionally, static compute metrics are poor indicators of capability as smaller, efficient models frequently outperform legacy systems.

  6. Google argues that isolated domestic rules risk stranding Australian companies and that Australia should anchor guardrails in global technical standards and provenance protocols.

    Isolated domestic rules risk stranding Australian technology companies, raising export costs, and delaying access to new technology. Australia should anchor guardrails in global technical standards, aligning organisational management with ISO/IEC 42001, adopting content provenance protocols like C2PA, and participating in international efforts like the Hiroshima AI Process.

Recommendations

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

  1. Foster domestic research hubs capable of attracting and retaining world-class talent.

    To maintain a competitive edge, Australia must foster domestic research hubs capable of attracting and retaining world-class talent.

  2. Deliver oversight through sector-specific regulatory frameworks that evaluate applications against non-AI baselines.

    Because AI is deployed across diverse industries, from agriculture and resources to healthcare and finance, risk profiles vary significantly. Oversight is therefore best delivered through sector-specific regulatory frameworks evaluating applications against non-AI baselines.

  3. Adopt a hub-and-spoke model with domain regulators retaining authority and a central technical advisory.

    Google recommends a 'Hub-and-Spoke' model. Domain regulators would maintain authority, supported by a central technical advisory. This bridges technical expertise gaps across government while also ensuring oversight.

  4. Tailor high-risk guardrails strictly to high-consequence uses, leaving everyday business applications unencumbered.

    High-risk guardrails should be strictly tailored to high-consequence use cases, leaving everyday business applications unencumbered.

  5. Align accountability with operational controls across the AI lifecycle: upstream developers provide technical documentation, evaluations and baseline safety benchmarks for foundation models; downstream deployers manage operational contexts, user controls and human oversight.

    Accountability must align with operational controls across the AI lifecycle: ● Upstream developers: Provide technical documentation, evaluations, and baseline safety benchmarks for foundation models. ● Downstream deployers: Manage operational contexts, deploy user controls, and maintain human oversight.

    A program corrected obvious text-recognition errors in this quote for display.

  6. Base governance of general purpose AI on capability assessments and red-teaming rather than static compute thresholds.

    General Purpose AI models are not inherently high-risk; their risk profile depends on real-world application. Additionally, static compute metrics are poor indicators of capability as smaller, efficient models frequently outperform legacy systems. Governance should focus on capability-based assessments and red-teaming benchmarks instead of static compute thresholds.

  7. Anchor guardrails in ISO/IEC 42001, C2PA provenance and the Hiroshima AI Process.

    Isolated domestic rules risk stranding Australian technology companies, raising export costs, and delaying access to new technology. Australia should anchor guardrails in global technical standards, aligning organisational management with ISO/IEC 42001, adopting content provenance protocols like C2PA, and participating in international efforts like the Hiroshima AI Process.

Read the original

In some quotes a program corrected obvious text-recognition errors for display.

Open in the interactive guide