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 557

Ms Jodie-Ann Jones

Submitter group: Individuals.

Position labels are not shown for individual submitters.

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.

An NDIS provider who built an AI safety layer with a provisional patent describes gaps from a compliance review, explicitly not a legal opinion. The submitter asks for vulnerable-user safeguards, Indigenous co-design, contestability, evidence and reporting standards, and trauma-informed AI Safety Institute expertise. They explicitly seek neither funding nor endorsement of their own work.

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 the people the Terms of Reference name as needing protection are least likely to be in the room for AI governance decisions.

    The elderly, people with disability, trauma survivors, and children are the people named explicitly in this Committee's Terms of Reference as needing protection from AI-enabled fraud, scams, and harm. They are also the people least likely to be in the room when AI governance decisions are made,

  2. Argues that there is no clear rule for when an AI system may lawfully override a user's own privacy setting in a safety-critical moment.

    No clear rule for when an AI system may lawfully override a user's own privacy setting in a safety-critical moment.

  3. Argues that there is no single national standard for how AI systems interact with state and territory mandatory reporting obligations.

    No single national standard for how an AI system should interact with mandatory reporting obligations. Those obligations differ by state and territory, and there is no equivalent national rule to build against when the system, not a human professional, is the one first receiving a disclosure.

  4. Argues that there is no admissibility standard for AI-generated or AI-logged records, and that a cryptographic hash alone does not establish reliability.

    No admissibility standard for AI-generated or AI-logged records. A cryptographic hash proves a record was not altered after it was created. It does not establish when or how it was collected, and does not by itself meet the reliability-of-process tests courts already apply to digital evidence

  5. Says general-purpose AI systems already in use are a live risk to vulnerable people, citing a 2026 Cornell Tech study.

    First, general-purpose AI systems already in everyday use are a live risk to this exact population. A 2026 Cornell Tech study found that Gemini and ChatGPT could be steered, through persona or settings manipulation, into producing content that facilitates coercive control against intimate partner violence victims.

Recommendations

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

  1. Issue statutory guidance on when and how AI may lawfully use safety exceptions to users’ privacy settings, requiring disclosure and logging every time.

    Issue clear statutory guidance on when and how an AI system may lawfully act on a safety exception to a user's own privacy settings, including a mandatory disclosure and logging standard for every time that exception is used.

  2. Establish a single national standard for AI handling of disclosures that would trigger mandatory reporting.

    Establish a single national standard for how an AI system must handle a disclosure that would trigger mandatory reporting, so developers are not left reconciling eight different state and territory regimes on their own.

  3. Set an evidentiary standard for AI-generated and AI-logged records, developed with the courts.

    Set an evidentiary standard specific to AI-generated and AI-logged records, developed with input from the courts, so that neither a vulnerable user nor a developer is left guessing whether an export is actually usable as evidence.

  4. Require co-design with an Aboriginal Community Controlled Organisation under CARE Principles as a certification or funding condition for any AI system claiming to serve Aboriginal and Torres Strait Islander users.

    Make co-design with an Aboriginal Community Controlled Organisation, in line with the CARE Principles, a condition of certification or funding for any AI system that claims to serve Aboriginal and Torres Strait Islander users — not a voluntary best-practice note.

  5. Require a contestability and human-oversight standard for automated decision-making in emotionally or psychologically sensitive AI tools, distinct from general AI ethics guidance.

    Require a contestability and human-oversight standard for automated decision-making in emotionally or psychologically sensitive AI tools specifically, distinct from general AI ethics guidance, given how different the stakes are when the "decision" is whether a human being is told someone may be at risk.

  6. Resource the AI Safety Institute with trauma-informed and emotionally-sensitive-design expertise.

    Resource the Australian AI Safety Institute with dedicated trauma-informed and emotionally-sensitive-design expertise (term of reference 13), not only technical and cyber safety expertise.

  7. Treat the gaps described in the submission as real.

    I am not asking this Committee to take my word for any of this. I am asking it to treat the gaps described above as real, because I found them by trying to do this properly, not by trying to get away with doing it poorly.

  8. Not let the law simply trust that harm to vulnerable people from AI will not happen again.

    If a sole trader, working alone from a home office in regional New South Wales, can identify this many specific points where the law has not yet caught up to this category of AI system, there will be developers who never check at all. The people who get hurt by that will be the same people I spent eighteen months trying to protect in an aged care facility that nobody in authority wanted to believe me about. I would ask the Committee not to let the law simply trust that it will not happen again.

Read the original

Open in the interactive guide