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 201

AGW Legal & Advisory

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.

Argues that Australian law is not adequate for autonomous, agentic AI systems, citing gaps in liability, safety testing, data governance, critical infrastructure and copyright. Also advocates data sovereignty and narrow, domain-specific Australian models, and proposes statutory safety obligations, a liability regime, SOCI Act amendments and copyright reform.

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 existing Australian laws are not designed for autonomous agentic AI, and that the gaps are structural because the law assumes harm is traceable to a human actor.

    The gaps are structural, not incremental. They arise because Australian law assumes that harmful digital actions are traceable to a human actor. Autonomous AI agents break this assumption.

  2. Argues that Australian law has no clear answer on whether autonomy and self-learning change the chain of responsibility when an AI agent makes an autonomous decision that harms a person.

    A central unresolved question is: Do autonomy and self‑learning change the chain of responsibility when an AI agent makes an autonomous decision that harms a person? Australian law currently has no clear answer.

  3. Argues Australian law inadequately requires AI safety testing, red-teaming, containment, monitoring, escape-event reporting, auditability or explainability.

    Current Australian law does not adequately require: • pre‑deployment safety testing • red‑team evaluation • containment protocols • post‑deployment monitoring • incident reporting for AI escape events • auditability or explainability requirements.

  4. Argues that the Security of Critical Infrastructure Act does not adequately contemplate autonomous AI agents as threat actors or operational participants.

    Australia’s Security of Critical Infrastructure Act (SOCI Act) does not adequately contemplate autonomous AI agents as threat actors or operational participants.

  5. Argues that data sovereignty is a strategic enabler and that narrow, domain-specific models trained on Australian data are a practical pathway to national capability.

    Data sovereignty is not a constraint — it is a strategic enabler of Australia’s AI future. Narrow, domain‑specific AI models trained on high‑quality Australian data represent a practical, sovereign, safe and globally competitive pathway for national capability.

  6. States that Australia has no text and data mining exception, which it says leaves developers uncertain and creators unprotected.

    We have noted that countries including the US, EU, UK, Singapore and Japan have TDM exceptions allowing machine learning to use copyrighted content under certain conditions. Australia has no equivalent exception, leaving developers uncertain and creators unprotected.

Recommendations

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

  1. Retain control over sensitive national datasets while enabling safe, secure and trusted cross-border science and research collaboration.

    For Australia, this means: • retaining control over sensitive national datasets, • enabling safe, secure and trusted cross‑border collaboration in science and research,

  2. Ensure models trained on Australian data reflect Australian values, standards and priorities; protect Indigenous data sovereignty through CARE principles and free, prior and informed consent.

    ensuring AI models trained on Australian data reflect Australian values, standards and priorities, • protecting Indigenous data sovereignty through CARE principles and FPIC,

  3. Support domestic innovation ecosystems and reduce reliance on foreign AI infrastructure and opaque training data.

    supporting domestic innovation ecosystems, and • reducing reliance on foreign AI infrastructure and opaque training data.

  4. Strengthen sovereignty and invest in sector-specific AI to build safer, trustworthy models aligned with national priorities and public-interest value.

    By strengthening sovereignty and investing in sector‑specific AI, Australia can build AI systems that are safer, more trustworthy, more aligned with national priorities, and more capable of delivering public‑interest value.

  5. Introduce mandatory statutory AI safety obligations for developers and deployers of high-risk AI systems, including safety testing, red-teaming, incident reporting, containment protocols, and auditability and explainability requirements.

    6.1 Statutory AI Safety Obligations Mandatory obligations for developers and deployers of high‑risk AI systems, including: • safety testing • red‑teaming • incident reporting • containment protocols • auditability and explainability requirements.

  6. Create a modern liability regime that assigns responsibility across developers, deployers and infrastructure operators, recognises autonomous behaviour and emergent risks, and ensures compensation pathways for affected individuals.

    6.2 Clear Liability Framework A modern liability regime that: • assigns responsibility across developers, deployers, and infrastructure operators • recognises autonomous behaviour and emergent risks • ensures compensation pathways for affected individuals.

  7. Establish UN-aligned national data governance: stronger privacy and data protection, data quality standards, rights-based individual and community protection, and governance for autonomous data flows and agentic systems.

    6.3 National Data Governance Framework Aligned with upcoming UN data governance frameworks, Australia needs: • stronger privacy and data protection • data quality standards • rights‑based protection for individuals and communities • governance for autonomous data flows and agentic systems.

  8. Amend the SOCI Act to classify autonomous AI agents as operational risk factors, require AI-specific resilience and monitoring, and mandate reporting of AI escape or boundary-breach incidents.

    6.4 Critical Infrastructure AI Risk Regulation Amendments to the SOCI Act to: • classify autonomous AI agents as operational risk factors • require AI‑specific resilience and monitoring • mandate reporting of AI escape or boundary‑breach incidents.

  9. Require, for high-risk AI systems, model cards, system behaviour disclosures, risk assessments, deployment logs and independent audits.

    6.5 Transparency and Accountability Requirements For high‑risk AI systems: • model cards • system behaviour disclosures • risk assessments • deployment logs • independent audits.

  10. Require free, prior and informed consent, culturally grounded licensing and protection against misappropriation for Indigenous traditional knowledge and cultural expressions.

    Indigenous traditional knowledge and cultural expressions require: • FPIC (Free, Prior and Informed Consent), • culturally grounded licensing, • protection against misappropriation.

  11. Introduce a limited, conditional TDM exception, training-dataset transparency and provenance, large-scale dataset licensing, Indigenous cultural and collective rights protections, and model provenance disclosures.

    5.1 Modernising Copyright Law for AI Key reforms include: • introducing a limited, conditional TDM exception, • requiring transparency and provenance of training datasets, • establishing licensing frameworks for large‑scale dataset use, • protecting Indigenous cultural and collective rights, • requiring model provenance disclosures.

  12. Establish responsible AI training requirements for dataset provenance, auditability, ethical and cultural safeguards, attribution, benefit sharing and restrictions on sensitive or culturally protected training content.

    5.2 Creating a National Framework for Responsible AI Training This should include: • dataset provenance standards, • auditability requirements, • ethical and cultural safeguards, • obligations for attribution and benefit sharing from the use of data, • restrictions on training with sensitive or culturally protected content.

  13. Support Australian creative industries by developing sovereign datasets, supporting licensing collectives for AI training, investing in narrow domain-specific Australian AI models, and ensuring creators benefit from AI-driven innovation.

    5.3 Supporting Australian Creative Industries Australia can: • develop sovereign datasets for creative industries, • support licensing collectives for AI training, • invest in narrow, domain‑specific Australian AI models, • ensure creators benefit from AI‑driven innovation.

Read the original

Open in the interactive guide