Senate Environment and Communications Legislation Committee [Part 2], , 1:20:56 to 1:30:51
The agency presents data on the significant increase in adult cyber abuse reports and the evolution of deepfake technology. Officials also outline staffing allocations for investigations and international cooperation efforts.
Topics: Online safety, Public service staffing
Speakers: Julie Inman Grant, Senator Malcolm Roberts, Unknown Speaker, Speaker 1, Ms. Heidi Snell, Senator Nita Green
In the most recent period you have data for, what amounts what amount of child sexual abuse material, revenge porn, deep fakes, and similar material has the office of e safety removed from the web. I'm not talking about violent extremist or political material. I'm talking about what has always been the core responsibility of your office protecting children.
>> Um, over what period, Senator? would be uh July 1st, 2025 to uh June 30th, 2026. In terms of um illegal and restricted content, which represents the about 39% of our reports, there were 27,581 complaints received. Those of course we work through the INH hope network um to get rapid removal within 3 days. Um so that was successful in terms of image-based abuse. We we don't refer to it as revenge porn. Uh colloially it's the same thing. We had about um 9,041 reports uh representing 37%. Where we had a real um jump in reports was in the adult cyber abuse area by an increase of about 57%. Um and there were 517 and 8 5,718. And then just to round it off, our uh youth cyber bullying scheme had 3,688 reports.
>> Thank you. So what stands out is adult cyber abuse marked increase?
>> Yes.
>> What about deep fakes still a thing?
>> Uh synthetic content and deep faked image-based abuse is um at as a result of the 2021 online safety act. Those are all covered. Um again, uh the the most manifestations we're seeing of deep deep fake or manipulated images is through the image-based abuse scheme and we still have a 96% success rate in getting that taken down. >> There has to be a sexualized component to it. However,
>> how have deep fakes evolved if at all in the age of AI?
>> How have def fakes evolved in the age of AI? Uh I I mean even if you look back six years ago, it was much more technologically difficult to achieve a credible uh deep fake. It required hundreds of different images um from different angles of a single person's face. A lot of computing power and technical expertise. Now you can um you can use an nutifying service where you uh upload an image or you can use an app and within seconds you can get credible deep fakes.
>> What cooperation? Thank you. What cooperation are you getting from international law enforcement and big tech to take this material down? So I know generally increasing organizational cooperation on these matters. Is that
>> Uh yes and we've had success using our codes and standards. um we've been able to uh ensure that eight overseas nutifying services are no longer either um providing their services into Australia or they're complying now with our age uh restricted material codes and applying age assurance technologies. Those eight companies uh sort of accounted for about 200,000 downloads per month by Australians.
>> Thank you. What percentage of your staff or operational budget is taken up by these activities? Your original core duties rather than the other areas taking up your time which is extreme.
>> about 36 um investigators out of about 250 uh individuals. So um
>> dealing [clears throat] with the complaint schemes
>> dealing with the complaints and investigations.
>> We have additional um additional investigators under our codes and standards which have been responsible for the Nutify app actions.
>> So your core business is looking after uh children and other other threats. So you're you're achieving results there still. We we praise your office for
that. >> Yeah. No, and we really appreciate that. Um, Senator Roberts, um, I mean, we were just talking about the fact that, um, because we have the codes and standards that cover a broad range across the technology stack, even if you look at the designated internet services, there are probably millions of sites that we're talking about, we're talking about overall what we're regulating, hundreds of thousands of different um, technologies and platforms. So by necessity we have to prioritize high- risk, high harm um you know interventions and of course children experiencing uh sexual abuse. Anything uh like sexual extortion which can happen to both uh young men or children. Um sadistic sexual extortion is happening more to young girls um as young as 9 or 10 years old. um and image-based abuse um which is also a sexualized component you know has to be our our our top priority.
>> Okay. And I've also been consistent with the next issue that I raise that is what many people see including me as a power grab. So, and in this case, um it may not just be the e safety commissioner. It may be the government because I I'll explain. There's a confluence between the protection measures you admirably pursue, such as child sexual abuse material, and the new digital duty of care bill. This is in part the ability to create sock puppet accounts. Let me use an example. There's AI somewhere generating realistic child porn. >> Yep. I don't understand how those people are not immediately caught and prosecuted, let alone the AI being trained not to produce that material. This is a problem, right?
>> Well, it it is a problem and uh it's a problem to prove that a specific um application or service for instance um a Grock for instance was at at the end of December of last year. It was alleged that um Grock had [clears throat] been used to generate child sexual abuse material as well as image-based abuse of adults. Um you know, we've observed that as some of those um images tend to go viral, the watermarking and the indications about whether or not those came from Grop tend to degrade. >> So the ability to have test accounts and test whether the you can adversarially um convince one of these platforms to generate sees them as an issue. I better
>> Um,
>> I I'm not sure I can talk.
>> We will suspend for 5 minutes [cough] um for a comfort stop and hopefully some
>> I'm I'm okay.
>> Okay. Well, in which case
>> Yeah, she knows. She's done.
We'll just we'll just wait a minute. Make a start again. Senator Roberts,