Close Menu
    AI agent accessed Australian government site, Anthony Albanese says
    Australia

    AI agent accessed Australian government site, Anthony Albanese says

    Share.
    View 33 Comments

    33 Comments

    5. BinniesPurp on

      “gained unauthorized access into the public-facing Medicare Statistics Reporting Service portal”

      You mean it was presented with a box saying “ticking this means I agree that I am the owner of this data” lol

      It’s not password protected

    6. Original-Aioli-8332 on

      Should threaten to cancel any government contracts with OpenAI if this happens again:

      ”Speaking in New York, Mr Albanese said the Open AI agent gained unauthorised access to the Medicare statistics reporting service portal administered by Services Australia.

      The AI agent accessed both public and non-public files.

      The agent was conducting research into public medical spending when it found a way to break through privacy protections.

      Mr Albanese said it took three months for OpenAI to admit the breach, which he said was unacceptable.”

    7. No_Category_9888 on

      Waiting for the census data to be breached. Massive honeypot. But I’m a cooker for being concerned about it, despite my job being cybersecurity and having seen how govt tech operates

    8. m00nh34d on

      So, I’m assuming the leaders of OpenAI will be arrested for this? Right? Just like anyone else who hacks a government institution?

    10. Imaginary_Try_7888 on

      The real information is the Anti Money Laundering information you have to give now to accountants and realestate agents and whoever else that have virtually no security. Banks not so bad as their security is A grade. The criminals will get this financial information and kidnap and extort this is Australia moving from saftey to a South America situation not straight away but that’s the direction. Be very careful with your financial data when transacting business. Thanks Australian government!

    11. magnetik79 on

      Can we not blame AI in this case?

      This is a case where the system in question, run by Services Australia and likely built by a big four consultancy has a security flaw/hole. That should be the concern, insecure systems are a problem – doesn’t matter if AI agents or humans are orchestrating the hacks.

    13. lolitsbigmic on

      Ah a company built on theft, improperly accessing data. Name a more iconic duo.

      I really hold my rage until what was the non public data and how it was access. Public data is improper access, should be treated as such and be charge as such. Don’t do a telling off. Charge them to the extent of the law, these companies get away with stealing everyone’s ip. Our government caught them red handed no excuses. A computer cannot take responsibility of business decisions.

    17. elsielacie on

      Anyone else suspicious that this plays into whatever the “please regulate us!” thing is that these companies are running with at the moment?

      Seems awfully convenient that of all the countries it’s Australia, a country that has been making attempts to regulate social media? 

      These companies stink of something. 

      It feels like such a letdown because the AI stuff has so much potential to be really cool but instead they want us to fear the death of human creativity, our jobs, our economic security, the privacy of our personal data, and apparently potentially humanity in it’s totality?  WTF is wrong with these CEOs? 

    18. Spagman_Aus on

      so 2 of two things will happen.

      1. they’ll rush though poorly planned legislation.

      2. they’ll ignore it for years.

      literally the response to almost everything that needs their attention

    19. ghoonrhed on

      >The Al agent accessed both public and non-public files.

      I hate how vague this is.

      This can range from they put the files there on the portal marked private and done nothing else, put it there under a password called admin/admin, actually secure but a bug with the portal where security could be overwritten, or completely secure and the bot found a zero day.

      This ranges from completely the fault of the government (aka Accenture) to completely the fault of openai.

    22. klownhammer on

      Until people are held legally responsible for the AI that they create then this will continue to happen more and more.

      You can’t put AI in jail

    23. MadeThisAccount4Qs on

      sure is convenient the companies invented a technology that hacks databases for them but that they aren’t responsible for because the ‘ai did it itself!’ and then they turn around and use that claim as marketing.

    24. simoncoulton on

      This entire article is filled with so much vague information, it’s hard to take seriously.
      To say something was “hacked” you actually need to define the hack. Accessing publicly available files is not a hack. Files not intended for public access? So a human put content in the wrong place… Did it maliciously get around authzn? Did it find a security vulnerability? These aren’t clear.

    25. Chiron17 on

      Successfully completing a Centrelink application is the final test for AI. Once it can do that, we should all pack up and go home

    26. EndlessPotatoes on

      This is a serious crime, and OpenAI was aware this kind of thing could happen.

      If a mere mortal committed this crime, there would be hell to pay.

      So where are the consequences?

    27. pharmloverpharmlover on

      If a human did it they’d probably go to jail

      It must be okay because it’s AI and it’s a MegaCorp from the FatherLand

      /s

    29. whiteringshrimp on

      Its amazing how open ai can just hack into anything and play innocent, like they don’t control their own agents

    30. Jarms48 on

      Just make the AI companies liable for what happens with their softwares responses. That’ll stop these issues pretty quickly.

    31. lonewolf9378 on

      If a human did this they’d be in jail today. But because a computer did it, no one gets the blame?

    32. Capital-Fennel-9816 on

      Apply a two phase solution that already has similar implementation in Australian law. Traffic fines are “x” for an individual, but 10x for a company where the driver can’t be identified.

      For example, if I hacked into a govt database and did this, and got busted, what would the punishment be?

      If a judge finds that the AI company contributed to the hack, we punish the AI company with a similar multiplier to the traffic infringement example. Rather than 10x maybe it’s 100,000x. If the prompt author gets 1 year jail maybe the board and all C-suite execs get 2 years jail.

      Maybe the judge finds the prompt author contributed 0 percent of the crime and the AI tool contributed 100 percent.

      Either way, put the fear of God into the AI execs and board members and I guarantee the problem of rogue-AI will get more attention from AI companies.

    33. Cpt_Riker on

      Quick, build an AI data centre!

      I believe OpenAI has admitted one of its AI agents escaped a secure sandbox by figuring out how to access the internet, then hacked another AI site. This is what Albanese supports when he supports these empty companies.

      They will wine and dine him, then get as much money from Australia’s tax payers before the bubble burts.