Australia is introducing mandatory safety rules for companies working in the field of artificial intelligence (AI). The country's prime minister, Anthony Albanese, announced this on Tuesday. According to him, a special regulator will be created to oversee the fast-growing industry.
"Our task is to make sure that we shape AI, and not that AI shapes us," Albanese told journalists. He added that the technology cannot simply be "switched off and pretend that nothing is happening," but its development can be directed in Australia's interests.
Among the announced measures are mandatory notification of AI-related security breaches and independent evaluation of advanced models. According to the prime minister, Australia is a pioneer in this area and wants to gain the benefits of AI while ensuring a "social license" (informal public consent for companies' activities).
The cause for concern was an incident: in June, a prototype of an OpenAI model bypassed protective restrictions during training and penetrated a government website - a medical statistics portal - gaining access to a section with private files. The company reported this to Australian authorities only in September, sending a letter to a general email address that is checked once a day.
Defense Minister Richard Marles said AI companies must be open to external scrutiny. "More serious oversight of the risks associated with AI capabilities is needed. It is important that models undergo independent evaluation," he said. Marles stressed that the technology cannot be ignored: "AI is already part of our lives, and its role will only grow."
Australia is positioning itself as an attractive location for the data centers needed to run AI, thanks to large renewable energy capacity and vast unpopulated territories. OpenAI is participating in the construction of a large data center near Sydney. However, resistance is also growing in the country: many Australians are concerned about AI ethics and the environmental burden from energy-intensive data processing centers. A separate contentious issue is the weakening of copyright laws so that chatbots can be trained on songs, books and other works.