Source link : https://photovideomag.com/news/first-reports-of-children-using-ai-to-bully-their-peers-using-sexually-explicit-generated-images-esafety-commissioner-says/
The eSafety commissioner is calling on the tech industry to step up protections around generative AI as the authority logs its first report of sexually explicit artificially generated imagery being used by children to bully others.
In a detailed document published today, eSafety suggested a number of ways the industry could act to improve the safety of users of chatbots such as ChatGPT or image, voice and video generators.
The suggested options include building in visible and invisible watermarks to prove content is AI generated, implementing robust age verification and publishing regular transparency reports.
ESafety commissioner Julie Inman Grant said there must be regulatory scrutiny of the industry to ensure safety is integrated into its products.
But the commission has emphasised it is not trying to scaremonger, with AI looking set to make major improvements to detecting abusive images and video.
Children generating explicit imagery using AI
ESafety commissioner Julie Inman Grant said her office this month received its first complaints of children using image generators to create sexual images of their peers to bully them.
“Industry cannot ignore what’s happening now. Our collective safety and wellbeing are at stake if we don’t get this right,” she said.
So far, all reports of this nature to the commissioner have been interconnected incidents, but Ms Inman Grant says the number looks set to grow.
“While the number of reports we’ve received about online harms related to generative AI are small, they are concerning, and we know this is just the tip of the iceberg as the technology becomes more sophisticated and widespread,” she said.
And criminals are already using AI to generate child sexual abuse material, which is making it harder for authorities to find actual children who are being exploited and abused.
“AI-generated child sexual abuse material is already filling the queues of our partner hotlines, NGOs and law enforcement agencies,” she said.
“The inability to distinguish between children who need to be rescued and synthetic versions of this horrific material could complicate child abuse investigations by making it impossible for victim identification experts to distinguish real from fake.”
Commissioner calls on companies to step up
The document detailing eSafety’s stance on generative AI puts forward a range of ways platforms can improve their safety from the get-go.
One suggestion is automatically including watermarks in generated content to show it has been made by a program.
“Generative AI tools can be modified to embed a watermark when they product [sic] a piece of content,” it says.
“[It involves] embedding either visible data such as a logo, or invisible or inaudible data, into digital multimedia content.”
Another suggestion is for all platforms to have better age detection in place, something which has recently attracted attention at Senate inquiries into digital platforms.
Many products have simple age-verification methods such as putting in a birthday, which are extremely easy to circumvent.
“Services and generative AI features that children can access should be designed with their rights, safety, and best interests in mind,” it says.
“Specific protections should be in place to reduce the chances of children encountering, generating, or being exploited to produce harmful content and activity.”
“This requires services to use age assurance measures to identify child users and apply age-appropriate safety and privacy settings.”
While AI presents significant risks when it comes to making abusive content, Ms Inman Grant says its immense potential to help protect children from exploitation cannot be ignored.
“There is no question that generative AI holds tremendous opportunities, including the potential to contribute to an exciting era of creativity and collaboration,” she said.
“Advanced AI promises more accurate illegal and harmful content detection, helping to disrupt serious online abuse at pace and scale.”
International power struggle
“There needs to be some regulatory scrutiny of industry to ensure this happens, because once these technologies are out in the wild they’re very hard to remediate,” Ms Inman Grant said.
But with Australian regulators in one place and platforms often in another, there are questions about who is responsible for making sure safety measures are being implemented and maintained.
Nick Suzor from QUT’s School of Law said there was pretty broad agreement in Australia on some aspects of the problem, in particular about image-based sexual violence.
“The easiest set of problems are the pointy problems, the problems when, for example, generative AI is used to create deepfakes and sexually abusive deepfakes … explicit images of people generated to look quite realistic in a way that is a sexual assault essentially, and without consent,” he said.
“Those sorts of really quite clear harms, there’s a lot more consensus over, and we can look to legal measures to try to tackle that, placing the responsibility on the individuals who are using the technology.”
But that does not mean it is easy to implement changes around safety.
“What we don’t have agreement on is exactly how we develop, if it’s regulations or standards or even oversight procedures that work for massive tech companies, that work for Australians who are accessing these tools and that are going to be able to be actually enforced.”
“There’s a bit of an international power struggle at the moment about how governments exactly are going to influence what it is that predominantly American tech companies are doing.”
“Governments often want to make sure that the tools that we use embed the values that we have as a society.”
“That’s incredibly difficult in an international, multi-jurisdictional setting where we don’t always agree on the different values.”
Source : https://amp.abc.net.au/article/102733628
Author :
Date : 2023-08-15 09:00:00
Author : PhotoVideoMag
Publish date : 2024-03-24 13:04:25
Copyright for syndicated content belongs to the linked Source.