OpenAI

Prime Minister Anthony Albanese says regulation of artificial intelligence is more important than ever in the wake of an Australian government hack by one of America's AI giants.

The PM, who's currently traveling in New York, announced on Thursday a rogue OpenAI agent hacked Medicare's Statistics Reporting Service portal in June, but the company took months to notify the Australian government of the breach.

Deputy PM Richard Marles said OpenAI discovered the hack in August, but only told the government about it last weekend using a generic Services Australia email address.

OpenAI boss Sam Altman said the company's own investigation said no patient records were accessed by the agent, but the PM said it's not on.

I also expressed my disappointment that it took the company way too long to inform the government of what had occurred and the, the, the nature of the way that that notification occurred as well was unacceptable.

The government is establishing a task force to review the incident, which may have impacted as many as three other systems.

She says that Canada has signed more than 20 economic security and defense agreements across five continents over the past year.

Now, an OpenAI model hacked an Australian ministry website earlier this year.

Now, this marks one of the first known AI attacks on a government database.

The model gained unauthorized access to files on a website for reporting healthcare statistics.

Australian Prime Minister Anthony Albanese says that he has spoken to OpenAI's leadership about the incident.

Today, I spoke with the CEO of OpenAI, Sam Altman, to express Australia's extreme concern about this incident, and I also expressed my disappointment that it took the company way too long to inform the government of what had occurred and the, the, the nature of the way that that notification occurred as well was unacceptable.

Australia's Anthony Albanese there.

He added that there was no evidence that the personal information of Australian citizens was compromised.

It is the first known incident of an AI agent hacking into a government website.

I mean, he's spoken to Sam Altman, the chief executive of OpenAI, to express his concern.

I mean, it's just, I just can't get, it's unbelievable.

I

Number five, uh, models are leaving messages for other models.

So one of those incidents where OpenAI's model managed to, uh, break out of its containment resulted in messages being left for subsequent models to learn how to do that.

That is where you start to get little bit weird in my mind.

I mean, still human-like, not human, still worrying, not catastrophic, but it is maybe that, that glimmer, that hint of the idea that these things are, uh, underestimated in our minds.

Hmm.

Eighteen, um, kinda linked to my previous one, which is we're starting to get decisions through the courts of who AI should call if and when a user appears to show some sort of dangerous signs in conversation with those LLMs that, as you've said, are AI.

So, uh, British Columbia has filed a lawsuit against OpenAI, basically blaming it, um, for not taking action, not warning people when someone expressed interest in launching effectively a, a, a mass murder attack, um, an attempted one at least.

And so feels like we're on the, uh, the cusp of this being decided in courts rather than being decided by the companies, which, um, I don't know which one I feel better about, whether the company should decide this or whether a, a judge should.

Yeah.

Yeah, I don't know how many titles OpenAI is going to win, but anyway.

Mate, thank you for joining us, mate.

My pleasure.

Meta CEO Mark Zuckerberg has an estimated fortune of $252.5 billion, while Nvidia CEO Wang is worth $198.1 billion.

Technological executives expected at the dinner include Microsoft CEO Sejah Nutella, Apple Executive Chairman Tim Cook, OpenAI CEO Sam Altman, and President Greg Brookman, and Google CEO Sundar Pichai.

The guest list extends beyond the technological sector.

JPMorgan Chase CEO Jamie Dimon, Citigroup CEO Jane Fraser, and Pfizer CEO Albert Varela are expected to be there, according to a Reuters News Agency report.

He pointed to a drone carrying explosives at Leipzig airport, which Germany blames on Moscow.

The heads of Anthropic and OpenAI have warned the UN Security Council about the dangers of artificial intelligence.

They called for international cooperation and said key decisions on AI should be made through democratic processes, not in laboratories in San Francisco.

And Harvey Weinstein has been sentenced to 15 years in prison by a New York court.

At least two people are reported to have been killed.

The heads of Anthropic and OpenAI have warned the UN Security Council about the dangers of artificial intelligence and have called for international cooperation.

President Trump hosts Xi Jinping for a state visit this afternoon with talks at the White House and a state dinner later.

And Harvey Weinstein has been sentenced to 15 years in prison by a New York court.

Because they're in this split mentality right now where they have...

OpenAI does the same thing, where you have the internet-hosted agentic loop But then you also have the Codex CLI and Cloud Code that identically, which is client-side.

And the better architecture is the client-side one.

It's not the centralized one.

No one can tell me that they didn't feel a difference if they use the model for more than just a Hello World or whatever, right?

Compared to what? Because I've never been into Cursor.

So the Composer model just feels like... a very capable engineer when you tell them clearly what to do they crank it out at a decent quality not a fantastic quality so you still need that guidance whereas like Kimi was always like very generalistic and like you and in a similar way also like from what you mentioned with the personality aspect you just need to look at like use any anthropic model and any open AI model Particularly the newer OpenAI models, me as a German, feel very much seen there.

It's very direct communication.