Submission 435
Ms Grace Roberts
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.
Summary
Read the original submission for the full text.
Argues that Australia gives too little weight to serious AI risks, especially catastrophic risks including loss of control, biorisk and cyber attacks. Calls for an AI Act covering general-purpose AI regulatory gaps, a properly resourced AI Safety Institute focused on frontier safety, mandatory incident reporting and international engagement, to enable future benefits safely.
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.
Argues that AI capabilities and their associated risks are increasing at exponential rates.
AI capabilities (and their associated risks) are increasing at exponential rates (explainer).
Says an MIT AI Risk Initiative study of 272 experts as finding that 18 of 24 AI risk domains carry at least a 10% probability of catastrophic outcomes over five years.
A study that interviewed 272 international experts on AI by MIT AI Risk Initiative found as their top takeaway that many risks could have catastrophic outcomes (18 of 24 AI risk domains carry greater than or equal to 10% probability of catastrophic outcomes over the next five years).
States concern that the Australian Government gives AI risks, especially catastrophic ones, less weight than economic opportunities.
I’m concerned there’s an uneven focus by the Australian government on economic opportunities from AI and a lack of weight on the serious risks that come with it.
Says a Good Ancestors survey of 64 experts in which 93% said current laws are inadequate for managing the risk of AI systems escaping human control.
I've read the new AI Legislation Stress Test from Good Ancestors, which surveyed 64 experts across AI, cybersecurity, and policy. Their finding is stark: 93% of experts say our current laws are inadequate for managing the risk that AI systems might escape human control.
Argues that AI risks include misuse by individuals or states, not only misaligned AI.
These aren’t just risks arising from misaligned AIs going rogue, but from ordinary or powerful individuals or states using AI to further their malicious goals.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Promptly put more attention and resources into reducing catastrophic AI risks, including the harmful capabilities of AGI and ASI.
I implore the Joint Select Committee on Artificial Intelligence to promptly put more attention and resources to reducing the catastrophic risks of AI, including the harmful capabilities of artificial general intelligence (AGI) and artificial superintelligence (ASI).
Enact an AI Act covering general-purpose AI risks that do not fit within existing regulators' boundaries.
An AI Act that covers general-purpose AI risks that don't fit neatly into existing regulators' boundaries.
Properly resource the Australian AI Safety Institute.
The Australian AI Safety Institute (AISI) to be properly resourced - the UK's equivalent has 100+ staff and £66 million annually, and Singapore invested 15 times more despite only having a population less than a quarter of the size of Australia's.
Ensure AISI leadership focuses on frontier AI safety risks, rather than only business opportunities and less severe risks.
Leadership at the AISI who're actually focused on frontier AI safety risks, not just business opportunities and less severe risks.
Introduce mandatory AI incident monitoring, reporting and response planning that includes catastrophic risks.
Mandatory AI Incident monitoring, reporting and response planning to include a focus on catastrophic risks.
Include a focus on catastrophic risks in international and regional engagement.
International and regional engagement to include a focus on catastrophic risks.
Ensure risk management approaches, AISI priorities and legislation can keep up and adapt as AI capabilities and risks evolve rapidly.
Additionally, as AI capabilities and risks evolve rapidly our risk management approaches, AISI priorities and legislation needs to be able to keep up and adapt.
Consider the 'Ten positions for Australia in the age of AI' prepared by Good Ancestors.
I encourage you to also consider Ten positions for Australia in the age of AI prepared by Good Ancestors.