New recommendations to curb harms of generative AI

New recommendations to curb harms of generative AI

eSafety’s new tech trends position statement on generative AI provides specific Safety by Design interventions that industry can adopt immediately to improve user safety and empowerment.

“If industry fails to systematically embed safety guardrails into generative AI from the outset, harms will only get worse as the technology becomes more widespread and sophisticated,” eSafety Commissioner Julie Inman Grant said.

“This month, we received our first reports of sexually explicit content generated by students using this technology to bully other students. That’s after reports of AI-generated child sexual abuse material and a small but growing number of distressing and increasingly realistic deepfake porn reports.

“The danger of generative AI is not the stuff of science fiction. Harms are already being unleashed, causing incalculable harm to some of our most vulnerable. Our colleagues in hotlines, NGOs and in law enforcement are also starting to see AI-generated child sexual abuse material being shared. 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.

“And it has long been a concern that AI is being trained on huge datasets whose balance, quality and provenance have not been established, reinforcing stereotypes and perpetuating discrimination.

“Industry cannot ignore what’s happening now. Our collective safety and wellbeing are at stake if we don’t get this right.”

Incorporating advice from respected domestic and international AI experts, the paper details a range of safety measures and interventions, such as:

•            appropriately resourced trust and safety teams

•            age-appropriate design supported by robust age-assurance measures

•            red-teaming and violet-teaming before deployment

•            routine stress tests with diverse teams to identify potential harms

•            informed consent measures for data collection and use

•            escalation pathways to engage with law enforcement, support services or illegal content hotlines, like eSafety

•            real-time support and reporting

•            regular evaluation and third-party audits.

“There is no question that generative AI holds tremendous opportunities, including the potential to contribute to an exciting era of creativity and collaboration. Advanced AI promises more accurate illegal and harmful content detection, helping to disrupt serious online abuse at pace and scale,” Ms Inman Grant said.

“Let’s learn from the era of ‘moving fast and breaking things’ and shift to a culture where safety is not sacrificed in favour of unfettered innovation or speed to market.  If these risks are not assessed and effective and robust guardrails integrated upfront, the rapid proliferation of harm escalates once released into the wild. Solely relying on post-facto regulation could result in a huge game of whack-a-troll.”

The tech trends position statement also sets out eSafety’s current tools and approaches to generative AI, which includes education for young people, parents and educators; reporting schemes for serious online abuse; transparency tools; and the status of mandatory codes and standards.

Recently registered codes will soon require some social media services to take proactive steps to detect and remove child sexual exploitation material. eSafety is also currently considering a revised Search Engine Code, which directly considers generative AI. Mandatory standards are also being developed for Relevant Electronic Services and Designated Internet Services.

“While our regulatory powers around online safety are the longest standing in the world, regulation can only take us so far.

“Online safety requires a coordinated, collaborative global effort by law enforcement agencies, regulators, NGOs, educators, community groups and the tech industry itself. Harnessing all the positives of generative AI, while engineering out the harms, requires a whole-society response.”

The position paper also recommends that users understand what personal information is being accessed from the open web when using generative AI, and then take steps to protect their data. More information on popular generative AI-enabled services (such as Bing, Google Bard, Chat GPT and GPT-4) can be found on the eSafety Guide.

This story originally appeared as a media release from The Office of the eSafety Commissioner.