Submission 394
Name withheld
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.
voluntary industry commitments are not an adequate substitute for binding, independently enforced standards — particularly regarding the generation of child sexual abuse material, deepfakes, and other AI-enabled harms.
I submit that explicit statutory clarity is needed to put beyond doubt that AI- generated or synthetic child sexual abuse material is treated identically under law to material depicting a real child, regardless of the generation method.
I submit that voluntary AI safety commitments from industry are insufficient, and that Australia should legislate binding obligations comparable to the EU AI Act’s prohibited-practices list (Article 5) and transparency obligations (Article 50), rather than relying on company discretion.
I submit that Australia should establish, or empower an existing body as, a dedicated AI safety regulator with the power to investigate, audit, and impose penalties scaled to a company’s global turnover
Mandatory independent conformity assessment for high-risk AI applications, rather than industry self-certification;
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.
An independent regulator with real investigative and enforcement power, insulated from both industry and government control.
I submit that voluntary AI safety commitments from industry are insufficient, and that Australia should legislate binding obligations comparable to the EU AI Act’s prohibited-practices list (Article 5) and transparency obligations (Article 50), rather than relying on company discretion.
I submit that Australia should establish, or empower an existing body as, a dedicated AI safety regulator with the power to investigate, audit, and impose penalties scaled to a company’s global turnover — mirroring the EU AI Act’s enforcement structure — rather than relying solely on complaints-based, after-the-fact regulation.
The generation, distribution, or possession of child sexual abuse material is prohibited regardless of whether the depicted subject is a real person — synthetic, animated, or AI-generated CSAM is treated identically to material depicting a real child under the law.
Alongside CSAM, the following are banned outright, not merely “discouraged”: manipulative or deceptive AI techniques that materially distort behavior and cause harm; exploitation of vulnerabilities based on age, disability, or socio-economic situation; social scoring by public authorities; untargeted scraping of facial images to build recognition databases;
emotion recognition in workplaces and schools (except for narrow medical/safety uses); and real - time public biometric surveillance for law enforcement, except under narrow, safeguarded exceptions.
People must be told when they are interacting with an AI system rather than a human, when content has been AI-generated or manipulated (deepfakes), and such content must carry machine-readable marking.
Providers of general-purpose AI models must document their models, publish training-content summaries, adopt clear copyright policies, and proactively address systemic risks — not simply respond after harm is discovered.
A permanent, adequately funded regulatory body —independent of the companies it regulates — holds supervision and enforcement power, including the ability to investigate, fine, and compel compliance.
Penalties must scale with company size (percentage of global turnover, not a flat fee) so they remain a genuine deterrent to the largest AI developers, not a cost of doing business.
High-risk systems (used in justice, migration, critical infrastructure, education, employment, law enforcement, and democratic processes) require independent conformity assessment before deployment, not a company’s own internal sign-off.
AI providers must train models against a public, auditable set of safety principles, and demonstrate the model resists harmful requests even without external prompting to do so — this should be a baseline legal requirement, not something left to each company’s discretion or marketing.
The public release of model weights specifically fine-tuned or modified to remove safety guardrails (e.g., “uncensored” variants marketed for generating illegal content) should carry the same liability as distributing the illegal content itself.
Because safety training can degrade, be fine-tuned away, or fail to anticipate new misuse patterns, high-risk models require periodic independent re-testing, not a single pre-launch approval.