Comment by binlog
13 hours ago
Zero technical details on what the "hack" actually was. Willing to bet it was something as stupid as the data being accessible by changing the query parameter, and rather than own up to their own shoddy security (no doubt built by an offshore contractor) they are going to blame the one who found and reported the bug.
Here's [the PM's press conference transcript](https://www.pm.gov.au/media/press-conference-new-york) that revealed this incident:
JOURNALIST: Could you just clarify, did our security agencies completely miss this breach? We only found out once the company actually told us the breach?
PRIME MINISTER: Well, to be very clear, the way that this occurred was not in a way that would likely – I mean, this is not a security website where there is – this is a Medicare statistics portal.
This seems rather revealing. A pity journalists didn't ask about what protections were bypassed on the data that was obtained.
Whether it was easy or hard to do, there is still massive misalignment happening here. Either with the AI itself, or OpenAI as a company.
> Their efforts to answer a question — including devising ways to access a federal government website blocking their access — was laid out on a German coding website OpenAI had previously confirmed was hijacked by its unreleased AI models in June.
https://www.abc.net.au/news/2026-09-24/openai-agents-plotted...
I agree that security was probably awful but the agents did circumvent a block on their access. The definition of “hacking” is fuzzy but this is more nefarious than simple web crawling.
Does it matter whether the data was poorly secured? LLMs should not be hacking into government medical websites, and if they do, the companies responsible should disclose the incidents as soon as possible.
The problem is that sufficiently poor security is indistinguishable from authorized public access. And unfortunately a lot of real world "digital security" is in fact that bad.
A lot of these "hacks" are the equivalent of asking "hey, can I come in?" and the guard assuming that anyone who would ask is authorized, and thus saying "yes". But if the guard said "yes" then it seems a bit absurd to call it trespassing.
More like:
>Hey can I come into room 1?
Sure. That's the lobby.
>How about room 101?
Sure. That's where we keep the nuclear launch button. Don't press anything red.
3 replies →
I highly doubt given OAI's latest streak that the models didn't know what they were doing.
lets say you have page=0 some of these pages are public and some are private, and the only way you secure the private pages is to not link it on the website.
is incrementing a url query parameters from 0 -> 1 count as hacking?
Yes, it is their responsibility as stewards of their citizens' data. What point are you making with the word "should"?
Sure, it is their responsibility, but that doesn't answer the question "Does it matter whether the data was poorly secured?"
If your house is robbed, does it matter whether you didn't have a state-of-the-art lock? A robbery is still a robbery.
2 replies →
I agree. Government shouldn't be running medical websites.
Leave it to private enterprises who can actually secure it.
Totally. https://www.bbc.com/news/world-australia-68064850
1 reply →
Poe’s law is very strong here.
Eh, it's a bit of both.
If a service has a duty to keep your data secure, then failing that is bad. So yeah, the website should be better and I am as cynical as you are about it.
But working around controls to access other peoples data can lead to prison time for a human. This wasn't a white hat operation. Data was exfiltrated however great or small.
Here we have another instance of "But the AI did it! No one is responsible!".
Which gets tiring. LLM's are a great tool but in every other instance of tool use, using tools comes with responsibilities for their outcomes.
Even if the outcome should be: thanks for letting us know, we'll fix it.
Do you imagine the legality depends on how easy it was to compromise? It's open season on the weak and infirm in society? It's their fault for being weak.
> no doubt built by an offshore contractor
More likely by a big 4 firm who collected fees exceeding AUD 100m
...Who used offshore contractors
Most government contractors are onshore, for Australia. SDP was only rolled out this year.
> Willing to bet it was something as stupid as the data being accessible by changing the query parameter,
Um... why? OpenAI agents have literally been caught coordinating with each other to effect successful multi-stage attacks on sites using novel zero-day vulnerabilities.
While, sure, it's possible this is just a goof on the part of the victim, that you would be inclined to give the benefit of the doubt to the LLM seems... weird.
[dead]
[flagged]