It has been confirmed that an OpenAI agent gained unauthorized access to an Australian Medicare database.
Medicare is Australia’s publicly funded universal health insurance scheme that provides coverage for all Australian citizens and, as such, its database contains comprehensive personal and medical details of the entire population.
The AI agent accessed both public and non-public data, but at this stage, there is no evidence that any individual personal information has been accessed. One of the worst aspects of this incident is that it took OpenAI three months to notify Australian authorities of the breach.
Apparently, the AI agent was undergoing a learning/training exercise during which it accessed several Australian websites and services looking for answers and statistics relating to Australia.
An OpenAI spokesperson has since stated that the company is “conducting an extensive review of misaligned model activity during training”.
- You also might like: 99.9% Chance AI Will Destroy Humanity?
AI Development Requires Immediate Oversight
Ironically, we recently published an article titled Artificial Intelligence: The Perils & Pitfalls in which I called out the lack of regulations governing AI:
Currently, regulation of AI is practically non-existent. The danger here is that the major companies involved with AI development do not care about whatever faults might lie within their products; they are only interested in getting the technology out there as soon as possible so they can begin recouping their massive outlay.
This Australian Medicare incident must surely prove once and for all that AI, and especially AI agents, require the implementation of official third-party oversight as a matter of urgency.
To add even more fuel to the fire, three leading lights in AI development — Anthropic’s Dario Amodei, OpenAI’s Sam Altman, and SpaceX’s Elon Musk — have all recently come out and called for a slowdown in the pace of AI development.
BOTTOM LINE:
This is a scary incident, not only because the AI agent performed an illegal intrusion, presumably outside of its mandate, but also because the agent’s human handlers obviously lost control.
Surely, it should not be good enough for the AI company to simply say… oops, sorry, it was the AI’s fault. If something similar were to be perpetrated by black hat hackers, jail terms would be in the offing.
One only hopes that governments around the world proactively heed these warning signs.
—
Thanks Jim
Of course there is always the positive in this in that it is one less security hole that we now know about exposing yet another flaw in the Australian Governments ability to keep it’s Citizens’ data safe – despite it’s constant claims about how safe our personal government data is supposed to be. Luckily for the Government in a way that it wasn’t a human hacker that found the flaw because otherwise right now Albanese and his Government ministers would be red-facedly trying to explain how it happened on their watch. and how much of our personal data was going to end up for sale on the Dark Web !
Also strange that Albanese only released the information AFTER everyone had completed the national Census which was almost completely online this year and when he had a chance to preen on the World stage. Convenient also that it just happened to be announced only days after the competing Ai companies went worldwide asking for special legal protections which would potentially result in the prevention of some of their future competitors being able to compete on a level playing field.
Of course I wholly agree that we need to control Ai rather than the other way round but it is either us in the West or China that will win the Ai race. Do we really think that China or North Korea will play nice by any rules that are created either here in Australia or anywhere else and not try to hack our systems with Ai then ? My message is that we had better get a better handle on Ai really quickly and start using it to fix our own security flaws here in Australia until a foreign power or a lone hacking group starts using Ai and, as usual, we are caught napping.
You can’t stop the progress of Ai – the genie is already out of the bottle and the sooner we realise that the better we can prepare for the future consequences.
Cheers
Reg