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.

Give the AI Safety Institute statutory powers, including access to models

8 submissions. Each submission counts once for this shared ask. Counts do not measure importance or merit.

A shared ask groups requests for substantially the same action from three or more submissions when the group was created. Groups remain listed if withdrawals reduce their counts. AI models summarised recommendations, grouped them and checked the groups. The label summarises the requests; it is not a quote. A shared ask does not mean agreement on anything else.

Theme: Australian AI Safety Institute

Recommendations from submissions

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. Extracts are chosen by an AI model. Submissions are listed in submission number order; recommendations follow their order on each submission’s page.

  1. UNSW AI Institute Submission 19

    1. Give the AI Safety Institute access to models, compute, data and controlled environments to independently evaluate and certify high-risk AI systems before deployment, with formal access to Australian academic expertise.

      Increase the funding, technical resources and authority of the Australian AI Safety Institute in line with comparable international bodies. Give it access to the models, compute, data and controlled environments needed to independently evaluate and certify high-risk AI systems before deployment, supported by formal access to Australian academic expertise.

  2. Alannah & Madeline Foundation Submission 55

    1. Attached paper: Give the AI Safety Institute investigative powers to test generative-AI production of child sexual abuse material, supported by a legal framework.

      The AI Safety Institute should be given specific investigative powers to test the capacity of generative AI systems to produce child sexual abuse material. Without the ability to test guardrails, it is impossible to evaluate whether platforms are meeting their obligations. The Oxford Internet Institute and similar bodies provide a model for what this research capacity could look like. However, it requires a legal framework to support it.

  3. Samantha Connor AM Submission 229

    1. Give the Australian AI Safety Institute statutory public sector oversight powers, including powers to audit systems, compel vendor documentation, inspect training and testing material, investigate incidents and publish findings.

      Give the Australian AI Safety Institute statutory public sector oversight powers, including powers to audit systems, compel vendor documentation, inspect training and testing material, investigate incidents and publish findings.

  4. Victorian Trades Hall Council (VTHC) Submission 293

    1. Give the AI Safety Institute EU AI Act-modelled enforcement powers to redesign AI risking public/workplace safety, collect information on high-risk systems' functions/workings and hold non-compliant organisations accountable.

      Ensure that the AI Safety Institute has the powers necessary to enforce the law and shape outcomes. In order to be effective, a broadened AI Safety Institute will need broad powers modelled on those provided for in the EU AI Act. The regulator must have the power to redesign AI systems where they pose a risk to public or workplace safety, collect information about the function and workings of high-risk systems and ensure accountability against organisations which are failing to comply with relevant obligations.

  5. Miss Kate Landry Submission 421

    1. Resource the AI Safety Institute to require access to frontier systems deployed in Australia, compel information from developers and deployers, and give its assessments consequences.

      Empower the Australian AI Safety Institute (item m) to do its job. Give it the power to require access to frontier systems deployed in Australia, to compel information from developers and deployers, and to have its assessments carry consequences. Then resource it to the level that job requires.

  6. Name withheld Submission 429

    1. Recommendation 2 - Statutory powers for the Australian AI Safety Institute. Place the Institute on a statutory footing with power to compel model access, conduct evaluations, and investigate incidents with an Australian nexus, including access to reasoning traces and logs.

  7. Ms Taina Perdriau Submission 479

    1. The Australian AI Safety Institute and relevant national-security agencies should have the expertise, independence, resources and legal authority to independently evaluate frontier models, investigate serious incidents, require rapid and transparent reporting of containment failures, and recommend delaying deployment where adequate safety cannot be demonstrated.

  8. Mr Craig Weniton Submission 502

    1. Establish the AI Safety Institute in legislation with a defined mandate, information-gathering and model-access powers, a guaranteed multi-year appropriation well above the current $29.9 million, independence in its published findings and an annual report to Parliament.

      Establish the Australian AI Safety Institute in legislation with a defined mandate, information-gathering and model-access powers, a guaranteed multi-year appropriation substantially above the current $29.9 million, independence in its published findings, and an annual report to Parliament.

Open in the interactive guide