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 30

Digital Rights Watch

Submitter group: Civil society.

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.

Argues that current regulation has not protected human rights and that AI should serve all Australians rather than large technology corporations. Says vendor-promoted AI benefits need an evidence-led approach. Calls for guardrails on government automated decision-making, privacy reform, consent and compensation for creative works used in models, and attention to unequal harms from facial recognition and deepfakes.

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. Argues that AI companies can move faster than the governments and regulators that oversee them, across the Australian economy.

    This situation is not unique to scribes, but is replicated across the Australian economy, where AI companies can move faster than the governments and regulators that oversee them.

  2. Says it is concerned that government automated decision-making tools will repeat the failures of Robodebt.

    DRW is concerned that rather than learning from Robodebt, the use of automated decision-making tools (ADM) will instead repeat the failures.

  3. Says government automated decision-making needs transparency over data inputs and decision points, with human discretion treated as paramount and timely, cost-effective review.

    We need transparency in how these systems are used, including the data inputs (and data excluded from inputs), and decision points. Human discretion has to be treated as paramount, with access to appropriate review in a timely and cost-effective manner.

  4. Argues that privacy reform should address the collection, use and storage of personal information rather than low-quality outputs after a model is built.

    Privacy law has the capacity to target reform at the source of the problem – the collection, use and storage of personal information – rather than addressing low-quality outputs after a model is built.

  5. Argues that AI corporations should be compelled to obtain consent from, and compensate, individuals for use of their creative work.

    AI corporations must be compelled to seek consent from, and pay compensation to, individuals for the use of their creative work in their models.

  6. Argues that AI harms fall more frequently and severely on already vulnerable groups.

    The harms of AI are not felt equally across Australian society. Rather, they are visited more frequently and more severely on those members of society who are already rendered vulnerable and by exacerbating existing inequalities.

Recommendations

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

  1. Follow an evidence-led approach promoting individual human rights and curiosity about how AI could improve society.

    We are sceptical of the claims of AI benefits that are regularly provided by companies selling AI products and services, and would like to see Australia following an evidence-led approach that promotes both the human rights of individuals as well as curiosity for how the technology might be used to better our society.

  2. Introduce strong government ADM guardrails, transparency about included and excluded data and decision points, paramount human discretion, and access to timely, cost-effective review.

    We urgently need strong guardrails for government use of ADM. We need transparency in how these systems are used, including the data inputs (and data excluded from inputs), and decision points. Human discretion has to be treated as paramount, with access to appropriate review in a timely and cost-effective manner.

  3. Give government ADM decision-makers an effective feedback loop to assess decisions without waiting for large-scale harm.

    Decision-makers need to have an effective feedback loop, so they can tell if they are getting it right without having to wait for large-scale human harm and a Royal Commission.

  4. Progress privacy reform.

    We urgently need to progress privacy reform, and we are grateful to see progress on this issue in recent weeks by the government.

  5. Strengthen consumer protection laws, including in respect of AI-generated scams and misleading AI-generated content.

    There are also improvements to be made in strengthening consumer protection laws, including in respect of AI generated scams and misleading AI generated content.

  6. Recognise that artistic creative agency is inherent in all Australians, so use of human-made works by AI systems cannot be reduced to licensing regimes and fees.

    we call on the committee to recognise that artistic creative agency is inherent in all Australians, not just those who earn a living from the creative industries. For this reason, the use of human-made artistic works by AI systems cannot be reduced simply to licensing regimes and fees.

  7. Compel AI corporations to seek consent from, and pay compensation to, individuals for use of their creative work in their models.

    AI corporations must be compelled to seek consent from, and pay compensation to, individuals for the use of their creative work in their models.

  8. Compel companies to remove model content lacking consent even if technically difficult, with non-compliance penalties making removal appealing.

    Companies must be compelled to remove from their models content for which they do not have consent, even if they find this to be an engineering challenge. Penalties for non-compliance need to be set such that removing the un-consented content is an appealing option for AI corporations.

  9. Embed human rights and gender-sensitive frameworks throughout the AI system lifecycle.

    Given the speed and scale of digital transformation, there is an urgent need to embed human rights and gender-sensitive frameworks throughout the AI system lifecycle: from conception and design through development, deployment, evaluation, and regulation.

Read the original

Open in the interactive guide