Search engine rules to ban AI-generated abuse images

Jennifer Dudley-Nicholson |

Search engines, including Google, will face a new set of rules in Australia to ban illegal content.
Search engines, including Google, will face a new set of rules in Australia to ban illegal content.

Google, Bing, Yahoo and other search engines will be required to prevent child sexual abuse and terrorist content appearing in search results under a code introduced to govern the industry.

The code will ban the companies’ generative artificial intelligence tools being used to produce deepfake versions of the offensive material in one of Australia’s first set of AI regulations.

The eSafety Commission launched the Internet Search Engine Services Code on Tuesday following months of negotiations with internet giants over the measures that had to be changed after the launch of generative AI tools. 

The code will come into place alongside five other online safety codes covering areas from social media to app stores and will include penalties of up to $780,000 a day for companies that fail to comply with its provisions. 

AI experts welcomed the code but said more restrictions and technological advances will be needed to stop the scourge of artificial “class one material” online.

eSafety Commissioner Julie Inman Grant said the search engine code, created under the Online Safety Act, was an important addition to stop the spread of the “worst of the worst” content from being widely seen or shared.

eSafety Commissioner Julie Inman Grant
eSafety Commissioner Julie Inman Grant says the search engine code closes key access gateways. (Bianca De Marchi/AAP PHOTOS)

“It helps ensure one of the key gateways to accessing material – through online search engines – is closed,” she said.

“It will target illegal content and conduct, with significant enforceable penalties if search engines fail to comply.”

The code dictates that search engines take “reasonable and proactive steps” to prevent public exposure to illegal content such as child abuse, pro-terrorism or extremely violent material, and provide tools to report instances of it.

The regulations also apply to “artificial intelligence features integrated into the search functionality that may be used to generate” illegal content – an addition Ms Ingram said had to be added to ensure it dealt with all relevant risks.

“The sudden and rapid rise of generative AI and subsequent announcements by Google and Bing that they would incorporate AI functionality into their internet search engine services all but rendered the original code drafted by industry obsolete,” she said. 

“What we’ve ended up with is a robust code that delivers broad protections for children.”

Search engines will also be required to publish annual reports into illegal material found and removed from their services.

University of NSW AI Institute chief scientist Toby Walsh said removing the most offensive material from search engines was an important step and preventing its creation using AI tools was equally vital.

“These tools are, sadly, being used to generate such offensive and, in many cases, illegal content,” he said

“Ever since generative AI tools became available, the (Australian Federal Police) have seen a significant uptick in the amount of such content… so it’s definitely a real challenge.”

Prof Walsh told AAP banning illegal AI-generated images would become easier after technological advances allowed content to be digitally watermarked but, until then, regulations were crucial to taking action against it. 

“(The code) doesn’t fix the problem because there are lots of other ways of accessing these tools … but it’s an obvious place to start,” he said. 

In addition to search engine regulations, the eSafety Commission is drafting rules for another two industry sectors, including online messaging services and photo storage. 

AAP