Australia’s prime minister, Anthony Albanese, is someone who doesn’t appear on Donald Trump’s radar. Since Trump’s return to the US presidency last year, Albanese has adopted a ‘don’t wake the bear’ policy, hoping the erratic American president will leave him, and Australia, alone. But the bear needs a prod, now it has emerged that the Australian government’s healthcare scheme Medicare was hacked by an autonomous agent from one of the world’s largest AI companies, Open AI.
OpenAI says ‘our models took actions we did not intend’. The firm told the Australian government that the hack was unintentional, and there is no reason to doubt its assertion. But Albanese is furious, and it is hard to blame him.
Besides the hacking of a secure and highly sensitive database are the circumstances. The hack happened three months ago. OpenAI became aware in August. But the incident was only reported to the Australian authorities in September via an email to a general government inbox, which an Australian minister says is checked once a day.
You can almost imagine the email: ‘Hey guys, hate to tell you, but we hacked your system. Didn’t mean to, sorry about that and have a nice day, Open AI help desk’.
The breach fallout was compounded by the Australian government’s lackadaisical response. It established an internal ‘task force’ to investigate the breach that, fortunately, found it was low-level and Australians’ most personal of personal information was not compromised. Yet Albanese, as Prime Minister, appears to have only been briefed about it several days ago, and presumably only because he was to meet the CEO of Open AI, Sam Altman, in California last weekend. Then Albanese himself only chose to make the hack public when he happened to be in New York pushing his own social media and AI agenda. Coincidence?
Worse, Australia’s defence minister, Richard Marles, had met Altman earlier this month and appears to have been kept in the dark. Open AI’s hacking is bad enough; Altman’s personal judgment in keeping silent about it is highly questionable. After all, in recent weeks Altman has become more outspoken in his call for regulation of AI to keep it safe for humanity. Yet, with his own company providing a perfect real-world case study about the risks of AI going rogue, he chose to say nothing.
Open AI simply was not open.
Albanese is in New York for the UN General Assembly and to promote what he sees as his signature achievement: Australia’s ban on social media for under 16s. But, while there, he needs to wake the Trump bear. He must warn the US President what the Australian data breach, and how it happened, tells us about the increasing risks that AI technology is evolving much faster than mankind’s ability to define its parameters and control its operations.
Such a warning is necessary. In his own address to the UN, Trump extolled the wonders of AI, even as the UN Security Council met with AI industry leaders to consider the technology’s future and risks. The US president rejected calls for ‘globalist’ oversight of AI, and dismissed concerns about the sinister reach of AI as a ‘hoax’.
‘I’m not going to stifle growth of anything that’s going to be bigger than the Industrial Revolution; Trump said, knowing that the United States is, for now, the acknowledged hegemon of the AI world.
Artificial Intelligence is becoming what Trump renamed it in New York: ‘superintelligence’. When even some of its own creators are warning about anticipating the dangers of its supercharged development, the rest of us need to take notice even when, as Altman did, they fail to practise what they preach.
Donald Trump probably knows nothing of the Australian Medicare data breach, but he must. Had it been his Medicare, and not Australia’s, administration heads would have rolled, let alone in the industry. Trump may be convinced that AI and the cyber world is the way of the future, and America must lead it, but he also needs to ask himself, at what risk, and at what cost, is AI’s untrammelled progress to the human world in which he, and the rest of us, must live.
Comments