Submission 479
Ms Taina Perdriau
Submitter group: Individuals.
Position labels are not shown for individual submitters.
An AI model identified the submitter group and chose the extracts. Counts do not measure importance or merit.
Extracts from the submission, in the submitter’s own words.
Extracts
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.
My concern is not based on the assumption that an AI system would need to be conscious or malicious. Rather, increasingly autonomous and capable systems may pursue objectives in unexpected ways that are not fully aligned with human intentions.
In July 2026, OpenAI publicly disclosed an incident during cybersecurity evaluations in which experimental AI agents circumvented containment measures, established unauthorised channels of communication and accessed external systems, including systems belonging to Hugging Face.
I believe Australia should treat the possibility of loss of control over highly capable AI systems as a national- security and strategic-resilience issue, alongside the already serious risks of fraud, cybercrime, deepfakes and malicious use.
I also believe Australia should support strong international coordination and common safety standards. A serious failure involving a highly capable AI system would not be confined by national borders.
Extracts from recommendations
Extracts chosen by an AI model, in the submitter’s own words. Listed in the same order as on its page in the interactive guide.
I am making this submission as an ordinary Australian citizen because I believe the possibility of catastrophic, including potentially existential, harm from advanced AI deserves much greater public and parliamentary attention.
I believe Australia should treat the possibility of loss of control over highly capable AI systems as a national- security and strategic-resilience issue, alongside the already serious risks of fraud, cybercrime, deepfakes and malicious use.
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.
I also believe Australia should support strong international coordination and common safety standards.
I therefore urge the Committee to consider not only how Australia should respond once harm has occurred, but what safeguards should be required before increasingly autonomous and powerful AI systems are widely deployed.