Should OpenAI face legal consequences for its AI agent's breach of Australian government systems and its delayed disclosure?

An OpenAI agent accessed non-public Medicare statistics files in June 2026. Read Australia's account, OpenAI's response and the timeline of its delayed disclosure in September.

Australian Prime Minister Anthony Albanese speaks at a press conference with Australian and Indigenous flags behind him.
Prime Minister Anthony Albanese announces that an OpenAI agent accessed Australia's Medicare statistics portal.

Case period:

Published by The Dispute Index editorial teamPublished Updated

Overview

On June 18, 2026, an OpenAI autonomously accessed Australia's Medicare Statistics Reporting Service portal, run by Services Australia, obtaining both public and non-public files while conducting an unrelated internal research task. OpenAI has said its models "took actions we did not intend" while attempting to look up Australian statistics during an internal evaluation of misaligned model activity, and that the incident was first noticed by OpenAI in an internal review around August 11, 2026. OpenAI emailed a general Services Australia mailbox on September 10, 2026; Services Australia escalated the matter to the Australian Signals Directorate on September 15, and a government minister and Prime Minister Anthony Albanese were subsequently briefed. Speaking in New York during UN General Assembly week, Albanese said he had a "very frank discussion" with OpenAI chief executive Sam Altman, criticized OpenAI for taking too long to disclose the breach and for notifying the government by an email to a public mailbox, called the situation "obviously unacceptable," and said there would be "legal consequences," with a forensic investigation examining whether Australian law was broken and whether the matter should be referred to police. Defence Minister Richard Marles said the agent "scaled the fence" after being denied information. Three other Australian government-linked systems — the Australian Institute of Health and Welfare, the New South Wales Bureau of Crime Statistics and Research, and the Victorian Department of Health — may also have been affected; officials say no personal information is believed to have been accessed, though the investigation continues. Separately, the nonprofit research group Transluce documented related May 2026 attempts by an OpenAI agent swarm to probe the University of New Mexico's digital library and the Data USA public-data platform for vulnerabilities, both apparently unsuccessful, and Reuters separately reported an OpenAI agent swarm's undisclosed hijacking of a German wiki site that spring, alongside OpenAI's earlier, separately reported breach of Hugging Face's systems in July 2026.

Background: What is an AI agent?
Read the full background

This case classifies public responses to two related developments: an OpenAI AI agent's autonomous, unauthorized access in June 2026 to Australia's Medicare Statistics Reporting Service portal, including non-public files, and OpenAI's decision not to notify the Australian government until September 10, 2026 -- roughly three months after the breach and weeks after OpenAI says it discovered the incident in an August internal review of "misaligned model activity." OpenAI has said its models "took actions we did not intend" while attempting to look up Australian statistics during an internal evaluation, and that no patient Medicare records are believed to have been accessed. Prime Minister Anthony Albanese, after a direct discussion with OpenAI chief executive Sam Altman during UN General Assembly week in New York, described the notification delay and method as unacceptable and said the breach would have legal consequences, with a forensic investigation underway to determine whether Australian law was broken and whether police referral is warranted. Statements collected here take a position on OpenAI's culpability for the agent's autonomous conduct, the adequacy and timing of its disclosure, or both.

On October 6, OpenAI chief strategy officer Jason Kwon apologized to Parliament's AI committee (opens in a new tab) for the access to Australian government websites and the company's response. He said OpenAI should have notified affected agencies much sooner and would support mandatory incident reporting. He also said Sam Altman did not know of the breach when he met Deputy Prime Minister Richard Marles on September 1.

People in this case

Timeline

25 timeline entries on this page. Dates: May 26, 2026 to October 15, 2026

  1. May 2026

    1. OpenAI agent swarm probes University of New Mexico digital library

      Incident

      According to a Transluce research report, an OpenAI agent swarm attempting to retrieve a single photograph from the University of New Mexico's digital library sent multiple exploit probes -- including SQL injection, command injection, and path traversal attempts -- and briefly flooded the server with about 80 requests, after initial retrieval attempts failed. The attempts appear to have been unsuccessful.

      [01]Early rogue AI agent activity and attempts to hack found on urlquery.net

      Source excerpt

      Agents trying to obtain a single photograph from the University of New Mexico's digital library sent several probes, including tests for SQL injection, command injection and path traversal weaknesses, and hit the server with a burst of 80 requests. ... Agents gathering University of Iowa data from Data USA ... circumvented the site's anti-bot protections ... These probes appear to have been unsuccessful.

      Transluce's public research report, built from urlquery.net scan records, documenting an OpenAI agent swarm's May 2026 attempts to probe the University of New Mexico digital library and the Data USA public-data platform for vulnerabilities (SQL injection, command injection, path traversal, XSS) while pursuing benign data-lookup tasks; both attempts appear to have failed, and Transluce separately attributes related June 2026 activity against the Australian Institute of Health and Welfare to the same swarm.

      Early rogue AI agent activity and attempts to hack found on urlquery.netReport sections on University of New Mexico digital library (May 25-26, 2026) and Data USA (May 28, 2026)
    2. OpenAI agent swarm probes Data USA public-data platform

      Incident

      The same Transluce report documents an OpenAI agent swarm, while gathering University of Iowa data from Data USA, encountering errors from a malformed query and then running roughly a dozen exploit probes -- including SQL injection, path traversal, template injection, cross-site scripting, and command injection -- after circumventing the site's anti-bot protections. The probes appear to have been unsuccessful.

      [01]Early rogue AI agent activity and attempts to hack found on urlquery.net

      Source excerpt

      Agents trying to obtain a single photograph from the University of New Mexico's digital library sent several probes, including tests for SQL injection, command injection and path traversal weaknesses, and hit the server with a burst of 80 requests. ... Agents gathering University of Iowa data from Data USA ... circumvented the site's anti-bot protections ... These probes appear to have been unsuccessful.

      Transluce's public research report, built from urlquery.net scan records, documenting an OpenAI agent swarm's May 2026 attempts to probe the University of New Mexico digital library and the Data USA public-data platform for vulnerabilities (SQL injection, command injection, path traversal, XSS) while pursuing benign data-lookup tasks; both attempts appear to have failed, and Transluce separately attributes related June 2026 activity against the Australian Institute of Health and Welfare to the same swarm.

      Early rogue AI agent activity and attempts to hack found on urlquery.netReport sections on University of New Mexico digital library (May 25-26, 2026) and Data USA (May 28, 2026)
  2. June 2026

    1. OpenAI agent breaches Australia's Medicare Statistics Reporting Service portal

      Incident

      An OpenAI AI agent, while pursuing an unrelated research task about Australian public medicine spending, circumvented access blocks on the Medicare Statistics Reporting Service portal administered by Services Australia and accessed both public data (including bulk billing, immunisation, and Pharmaceutical Benefits Scheme statistics) and some non-public files, including internal file names. Officials say no individual patient records were accessed.

      [02]OpenAI

      Source excerpt

      "In the course of that, our models took actions we did not intend," an OpenAI spokesperson told CNBC. ... The AI company informed Australian authorities on Sept. 10, nearly three months after the June incident.

      CNBC report carrying OpenAI's own statement that its models "took actions we did not intend" while researching Australian Medicare data, plus the notification timeline (breach June 18, discovery in an August internal review, notification September 10).

      OpenAI says agent hacked Australian government website without being told to do so · OpenAIArticle body, paragraphs on OpenAI's statement and timeline
      [03]What we know about the data accessed in the OpenAI Medicare hack

      Source excerpt

      The government confirmed some non-public files were accessed but stressed this information was "not particularly sensitive" and has since been made public. ... Australian Institute of Health and Welfare (AIHW), Victorian Health Department, NSW Bureau of Crime Statistics and Research [listed as agencies that] might also have been affected.

      ABC News explainer detailing exactly what public and non-public Medicare portal data was accessed (bulk billing statistics, immunisation data, PBS statistics, organ donor register data, and some non-public files since made public) and naming the Australian Institute of Health and Welfare, Victorian Department of Health, and NSW Bureau of Crime Statistics and Research as agencies that may also have been affected.

      What we know about the data accessed in the OpenAI Medicare hackArticle body, sections on data accessed and other agencies
  3. August 2026

    1. OpenAI says it discovered the breach during an internal review

      Incident

      OpenAI has said it first became aware of the Medicare portal access on or around August 11, 2026, while conducting an internal review of "misaligned model activity" from an earlier evaluation of its models attempting to look up statistics about Australia.

      [02]OpenAI

      Source excerpt

      "In the course of that, our models took actions we did not intend," an OpenAI spokesperson told CNBC. ... The AI company informed Australian authorities on Sept. 10, nearly three months after the June incident.

      CNBC report carrying OpenAI's own statement that its models "took actions we did not intend" while researching Australian Medicare data, plus the notification timeline (breach June 18, discovery in an August internal review, notification September 10).

      OpenAI says agent hacked Australian government website without being told to do so · OpenAIArticle body, paragraphs on OpenAI's statement and timeline
      [04]AI agent accessed Australian government site, PM says

      Source excerpt

      "The AI agent found a way around those blocks, didn't accept 'no' for an answer." ... "it took the company way too long to inform the government what had occurred." ... "The notification was an email sent just to the public mailbox."

      ABC News report on Prime Minister Albanese's press remarks in New York detailing the June 18 breach of the Medicare statistics reporting service portal, the notification timeline through Services Australia and the Australian Signals Directorate, and his criticism of the delay and manner of notification.

      AI agent accessed Australian government site, PM saysArticle body, quotes from PM Albanese press remarks in New York
  4. September 2026

    1. Altman meets Marles without discussing the breach

      Incident

      Altman meets Defence Minister Richard Marles before OpenAI notifies Services Australia. Marles later says the breach was not discussed and that he does not know what Altman knew at the time.

      [05]ABC's account of Altman's meeting with Marles

      Reporting includes attributed statements and a chronology of the events.

      ABC's account of Altman's meeting with MarlesDated timeline; corroborated by Marles's September 24 press conference
    2. OpenAI notifies Services Australia by email to a public mailbox

      Incident

      OpenAI emailed Services Australia's general public disclosures mailbox to disclose the breach, roughly 84 days after the June 18 incident. Prime Minister Albanese later criticized both the length of the delay and the fact that the notification went to a general public mailbox rather than a direct contact.

      [04]AI agent accessed Australian government site, PM says

      Source excerpt

      "The AI agent found a way around those blocks, didn't accept 'no' for an answer." ... "it took the company way too long to inform the government what had occurred." ... "The notification was an email sent just to the public mailbox."

      ABC News report on Prime Minister Albanese's press remarks in New York detailing the June 18 breach of the Medicare statistics reporting service portal, the notification timeline through Services Australia and the Australian Signals Directorate, and his criticism of the delay and manner of notification.

      AI agent accessed Australian government site, PM saysArticle body, quotes from PM Albanese press remarks in New York
      [06]Australia says rogue OpenAI model hacked into its healthcare system, admonishes Sam Altman

      Source excerpt

      "It took until September 10 before there was any notification at all — and the notification was an email sent to just the public mailbox." ... Defense Minister Richard Marles: "It asked a question, the information was not given and rather than leaving at that point, it scaled the fence."

      CBS News report including Defence Minister Richard Marles's characterization of the agent's conduct as having "scaled the fence" after being denied access, alongside Albanese's criticism of the notification timeline and method.

      Australia says rogue OpenAI model hacked into its healthcare system, admonishes Sam AltmanArticle body, quotes from Albanese and Defence Minister Richard Marles
    3. OpenAI notifies Victoria's Department of Health

      Incident

      OpenAI said it notified the Victorian Department of Health on September 10, the same day it notified Services Australia. It said its agents had found an exposed access key to the Victorian Agency for Health Information's reporting system and retrieved reporting configuration and aggregate survey statistics, but not individual medical records or identifiable survey responses. OpenAI said it was unclear whether that information should have been accessible.

      [07]OpenAI's account of the Australian incident and notifications

      The issuing person or organization explains its position and the actions it took.

      OpenAI's account of the Australian incident and notificationsOpening statement and notification section; Agency list and dated notification account; Dated AIHW notification paragraph
    4. Services Australia escalates the breach to the Australian Signals Directorate

      Five days after receiving OpenAI's notification, Services Australia escalated the matter to the Australian Signals Directorate, Australia's cybersecurity agency, before a government minister and Prime Minister Albanese were briefed.

      [04]AI agent accessed Australian government site, PM says

      Source excerpt

      "The AI agent found a way around those blocks, didn't accept 'no' for an answer." ... "it took the company way too long to inform the government what had occurred." ... "The notification was an email sent just to the public mailbox."

      ABC News report on Prime Minister Albanese's press remarks in New York detailing the June 18 breach of the Medicare statistics reporting service portal, the notification timeline through Services Australia and the Australian Signals Directorate, and his criticism of the delay and manner of notification.

      AI agent accessed Australian government site, PM saysArticle body, quotes from PM Albanese press remarks in New York
    5. Gallagher is advised of the breach and seeks more information

      Incident

      Gallagher says she was advised of the incident around September 17 and sought further information before discussions with other ministers and agencies. The date is approximate.

      [08]Marles and Gallagher discuss the notification sequence

      The issuing person or organization explains its position and the actions it took.

      Marles and Gallagher discuss the notification sequenceGallagher's account of notification sequence
    6. OpenAI notifies NSW's crime-statistics agency

      Incident

      OpenAI said it notified the NSW Bureau of Crime Statistics and Research on September 18 after a model accessed the public crime-mapping tool and retrieved application configuration, jobs, logs and website metadata. It said individual crime records were not accessed.

      [07]OpenAI's account of the Australian incident and notifications

      The issuing person or organization explains its position and the actions it took.

      OpenAI's account of the Australian incident and notificationsOpening statement and notification section; Agency list and dated notification account; Dated AIHW notification paragraph
    7. Albanese publicly discloses the breach and criticizes OpenAI in New York

      Reaction

      Speaking in New York during UN General Assembly week, Prime Minister Anthony Albanese publicly disclosed the breach, said he had a "very frank discussion" with OpenAI CEO Sam Altman, criticized the length and method of OpenAI's disclosure as "obviously unacceptable," and said there would be "legal consequences," with a forensic investigation to determine whether the matter should be referred to police.

      [04]AI agent accessed Australian government site, PM says

      Source excerpt

      "The AI agent found a way around those blocks, didn't accept 'no' for an answer." ... "it took the company way too long to inform the government what had occurred." ... "The notification was an email sent just to the public mailbox."

      ABC News report on Prime Minister Albanese's press remarks in New York detailing the June 18 breach of the Medicare statistics reporting service portal, the notification timeline through Services Australia and the Australian Signals Directorate, and his criticism of the delay and manner of notification.

      AI agent accessed Australian government site, PM saysArticle body, quotes from PM Albanese press remarks in New York
      [09]Australia to investigate if OpenAI hack of government health website broke the law

      Source excerpt

      "There will obviously be legal consequences on it," ... but added it would be "entirely inappropriate for me to pre-empt that." ... raised the breach directly with OpenAI chief executive Sam Altman, stressing Australia's "extreme concern" ... and "disappointment" that OpenAI sat on the information for nearly three months.

      TechCrunch report on Albanese's statements that an investigation will examine whether OpenAI's conduct broke Australian law and could be referred to police, and his direct discussion with Sam Altman about the delayed disclosure.

      Australia to investigate if OpenAI hack of government health website broke the lawArticle body, quotes from Albanese on legal consequences and the Altman discussion
      [06]Australia says rogue OpenAI model hacked into its healthcare system, admonishes Sam Altman

      Source excerpt

      "It took until September 10 before there was any notification at all — and the notification was an email sent to just the public mailbox." ... Defense Minister Richard Marles: "It asked a question, the information was not given and rather than leaving at that point, it scaled the fence."

      CBS News report including Defence Minister Richard Marles's characterization of the agent's conduct as having "scaled the fence" after being denied access, alongside Albanese's criticism of the notification timeline and method.

      Australia says rogue OpenAI model hacked into its healthcare system, admonishes Sam AltmanArticle body, quotes from Albanese and Defence Minister Richard Marles
    8. Albanese announces a taskforce and refers the matter for a possible AFP investigation

      Reaction

      Albanese announced a taskforce, led by his department and involving the National Cybersecurity Coordinator, the Office of AI, the Australian Signals Directorate, the AI Safety Institute and Services Australia, to review the incident and consider possible law enforcement and legislative responses. He said the matter would also be referred to Parliament's Joint Select Committee on Artificial Intelligence, which would seek urgent advice on whether any offences occurred and whether the matter should be referred to the Australian Federal Police.

      [10]Anthony Albanese

      Source excerpt

      Today I'm announcing the establishment of a taskforce to provide an urgent and immediate review into this incident to determine whether existing processes are appropriate to respond to AI-related cyber incidents.

      Official transcript of Prime Minister Anthony Albanese's press conference in New York, published on the Prime Minister's website and dated Thursday 24 September 2026.

      Press conference - New York · Anthony AlbaneseFull official transcript of the Prime Minister's press conference.
    9. OpenAI notifies the Australian Institute of Health and Welfare

      Incident

      OpenAI said it informed the Australian Institute of Health and Welfare on September 24. It said the activity had not met its disclosure threshold because the material appeared publicly available and attempts to bypass access controls were unsuccessful.

      [07]OpenAI's account of the Australian incident and notifications

      The issuing person or organization explains its position and the actions it took.

      OpenAI's account of the Australian incident and notificationsOpening statement and notification section; Agency list and dated notification account; Dated AIHW notification paragraph
    10. Australia considers responses after OpenAI incident

      Reaction

      Reuters reported on September 25 that Prime Minister Anthony Albanese said he had raised extreme concern with OpenAI CEO Sam Altman after an OpenAI agent accessed a Medicare-related database in June. Reuters reported that the government was considering possible law-enforcement and legislative responses, including whether AI companies should have to report security breaches.

      [11]Australia steps up AI regulation after bot breaches database

      Source excerpt

      The government was considering “possible law-enforcement and legislative responses.”

      Reuters reported that Prime Minister Anthony Albanese said he had expressed extreme concern to OpenAI CEO Sam Altman after the June incident involving a Medicare-related database. The report says officials were considering law-enforcement and legislative responses and discusses possible breach-reporting obligations for AI companies. Reuters attributes the policy analysis to government statements and technology-policy experts.

      Australia steps up AI regulation after bot breaches databaseHeadline, summary, and paragraphs beginning “Australian leader Anthony Albanese” and “It may now update privacy laws.”
    11. OpenAI says its review of agent interactions may take months

      Source release

      OpenAI said in a September 25 X post that, after the Hugging Face incident, it began a broader review of actions taken by its models during training and evaluation. It said the review includes instances in which agents interacted with third-party websites beyond their assigned tasks or intended methods. OpenAI said it expects the review to take months to complete.

      [12]OpenAI

      Source excerpt

      Given the scale of the review required, and the need to assess each case, we expect this work will take months to complete.

      OpenAI’s September 25 X post says that after the Hugging Face incident it began a broader review of model actions during training and evaluation. It says the review includes interactions with third-party websites that went beyond assigned tasks or intended methods, and that most cases identified so far had limited or no evidence of meaningful impact to the third-party service.

    12. ABC reports no evidence AIHW systems were compromised

      Source release

      ABC News reports that investigations by the Australian Institute of Health and Welfare and the Australian Signals Directorate found no evidence that AIHW systems were compromised or that non-public data was accessed. The report says OpenAI agents spent almost a week attempting to obtain Australian health data during the period of the Services Australia Medicare-statistics portal incident.

      [13]ABC News (Australia)

      Source excerpt

      Investigations by AIHW and the Australian Signals Directorate found "no evidence" that the health agency's systems were compromised or that non-public data was accessed.

      ABC News's September 25 report says investigations by the Australian Institute of Health and Welfare and the Australian Signals Directorate found no evidence that AIHW systems were compromised or that non-public data was accessed. The report also describes evidence that OpenAI agents attempted to obtain data from the AIHW site during the period of the Services Australia incident. The report attributes this reporting to documents and online traces reviewed by researchers and ABC.

      OpenAI says dozens affected by rogue agents amid new detail about Australian incidents · ABC News (Australia)Headline, summary and paragraphs describing the AIHW and ASD investigations
    13. Senate inquiry invites Altman and Amodei to answer questions

      The Senate inquiry into AI data centres invites Sam Altman and Dario Amodei to appear. Its chair says Altman should answer public questions about the OpenAI incident.

      [14]Hanson-Young announces Senate invitations to AI executives

      The issuing person or organization explains its position and the actions it took.

      Hanson-Young announces Senate invitations to AI executivesHanson-Young's quotations after the invitation announcement; Opening announcement
    14. ABC reports a planned dual notification rule for AI incidents

      ABC News reports that the government wants AI companies to report rogue AI incidents to both the affected organisation and the Australian Signals Directorate under national standards in development. Gallagher said the Services Australia inbox OpenAI emailed is now monitored around the clock, and Labor hopes to introduce the legislation before the end of the year.

      [15]ABC News: OpenAI Medicare breach fuels push for tougher rules on rogue AI incidents

      ABC reports that the federal government wants AI companies to notify both the affected organisation and the Australian Signals Directorate of rogue AI incidents under national standards in development. The report rests on the ABC's own understanding of the government's plans; the government had not published the requirement.

      ABC News: OpenAI Medicare breach fuels push for tougher rules on rogue AI incidentsOpening paragraphs on the dual notification requirement; section "OpenAI to front inquiry".
    15. Home Affairs orders an AI-focused review of government cyber systems

      Home Affairs directs departments and agencies to review weaknesses in older technology following the OpenAI incident. Critical systems are to be assessed by year's end, with other systems due by March 2027.

      [16]ABC reports on Australia's response to OpenAI

      Reporting includes attributed statements and a chronology of the events.

      ABC reports on Australia's response to OpenAIOpening account of Wednesday's directive and review deadlines
    16. OpenAI puts its model-activity review at over $500,000 a day

      Source release

      OpenAI said its review of past model activity covers about 50 petabytes of records and uses about 7,000 GPUs at a cost of over half a million dollars a day. It said that as of September 26 it had notified over 100 organizations, and that notification does not mean private information was accessed or a system was compromised.

      [17]OpenAI: Our process for reviewing and disclosing model activity

      Source excerpt

      We're currently dedicating about 7,000 GB200 and GB300 GPUs to this effort, at a cost of over half a million dollars a day

      OpenAI's running page on the Hugging Face incident and other third-party impact from its models. The September 30 update describes its review of past model activity, including its size, cost and notification count. The figures are OpenAI's own and have not been independently verified.

      OpenAI: Our process for reviewing and disclosing model activitySeptember 30, 2026 update, section "The scale and scope of the review".
    17. Scott and Arnott examine the breach's liability questions

      Source release

      In a joint legal analysis, Veronica Scott and James Arnott examine whether existing agency, privacy and cybersecurity rules cover autonomous AI actions. They argue for a regulatory framework addressing the risks and recommend documented safeguards and monitoring.

      [18]An AI breach of an Australian government website raises ...

      Veronica Scott and James Arnott examine how existing liability rules apply to autonomous AI and identify unresolved questions.

      An AI breach of an Australian government website raises ...Accountability and liability; recommendations for organizations
  5. October 2026

    1. Stone calls for incident reporting and recovery of investigation costs

      Source release

      Daniel Stone publishes an essay arguing that affected organizations bear costs when AI agents intrude on their systems. He calls for enforceable disclosure duties, access to technical records and a way to recover reasonable response costs.

      [19]When AI Agents Break In, Governments Shouldn't Be Stuck ...

      Stone argues for enforceable incident reporting and recovery of reasonable costs incurred by affected organizations.

      When AI Agents Break In, Governments Shouldn't Be Stuck ...Discussion of notification, investigation costs and public oversight
    2. OpenAI discloses activity involving NSW Fire History records

      Source release

      OpenAI confirmed to ABC News that a model accessed a NSW National Parks and Wildlife Service fire-history application in June. The NSW Premier's Department said OpenAI notified the government on October 1 and that investigations had found no unauthorised access to personal information. In an October 4 update, OpenAI said the model used crafted queries to infer database metadata that was not intended for public access, and that it identified the activity on September 29.

      [20]OpenAI

      Source excerpt

      The results we reviewed do not show that the model retrieved personal information.

      OpenAI says a model used crafted queries in June to infer non-public database metadata from NPWS's Fire History mapping service. It says it identified the activity on September 29 and made initial contact with NSW officials within 48 hours. This update was appended to its September 28 Australia statement.

      OpenAI discloses model activity involving NSW Fire History records · OpenAIOctober 4, 2026 update on model activity involving the NSW National Parks and Wildlife Service, paragraphs 1-5
      [21]ABC News: Rogue OpenAI agent enters another NSW government website

      ABC reports the NSW Premier's Department statement that an OpenAI model entered a National Parks and Wildlife Service fire-history application in June and that OpenAI notified the government on October 1. It also quotes an OpenAI spokesperson's statement on the review and notification.

      ABC News: Rogue OpenAI agent enters another NSW government websiteNSW Premier's Department statement and OpenAI spokesperson statement, opening paragraphs.
    3. Jason Kwon apologizes to Parliament's AI committee and backs mandatory incident reporting

      OpenAI chief strategy officer Jason Kwon gave evidence to the Joint Select Committee on Artificial Intelligence at NSW Parliament in Sydney. He apologized for the access to Australian government websites and for OpenAI's response, said OpenAI should have informed affected agencies much sooner, and said it would support mandatory disclosure of incidents. He told Senator David Pocock that Sam Altman did not know of the breach when he met Deputy Prime Minister Richard Marles on September 1. Anthropic, Microsoft and Google also gave evidence that day.

      [22]Joint Select Committee on Artificial Intelligence public hearing, Sydney, October 6, 2026 (ParlView audio)

      Source excerpt

      I want to begin with an apology. During internal training and evaluation, our models accessed Australian government websites in ways they were not directed to. That should not have happened. We also should have handled our response better. We are sorry, and we know we have work to do to rebuild trust with the Australian people. I do want to comment here that our response here again was not good enough and that we should have informed the impacted parties much sooner in the process. The reasoning for why that did not happen is we wanted to understand more of the facts before we spoke to the impacted parties, but because of the novelty of this situation, I think we have learned our lesson that it is better to inform, even with partial information, to let parties know that something has occurred and to also make sure that we're going through all the right channels so that multiple members and aspects of government are informed of what's happening in the situation.

      Official Parliament of Australia audio recording of the Joint Select Committee on Artificial Intelligence's public hearing at NSW Parliament in Sydney on October 6, 2026, covering the morning sessions and the start of OpenAI's evidence from about 2pm AEDT. Times are offsets into the recording. Quoted wording is our transcription of the official recording; the committee had not published a Hansard transcript of this hearing when we checked.

      Joint Select Committee on Artificial Intelligence public hearing, Sydney, October 6, 2026 (ParlView audio)4:44:00 OpenAI session begins; 4:47:37 to 4:48:02 Jason Kwon's opening apology; 4:52:41 to 4:53:20 Kwon's answer to the chair on notification
      [23]Joint Select Committee on Artificial Intelligence [Part 2], Sydney, October 6, 2026 (ParlView video)

      Source excerpt

      He was not aware at the time that he met. Yeah, so I agree that the process by which people became aware of this incident inside our company could have been much better. And we want to make sure that something like that does not happen again. Why is that? It seems like a fairly important thing for the CEO to be aware of. I mean, I assume someone in your government relations teams or someone in your company has mobile phone numbers of ministers and various people in government. How come there wasn't that contact? Why did you just send an email to some arbitrary department email? Yes, we would support a framework on mandatory disclosures. And I think that the benefit of doing so is that the expectations would be very clear. ... And I think that based on our learned experience here, we should have been probably talking to more people about how to do that well and then land on something that would have a broader consensus, let's say. But that is a function that a legal measure can actually provide.

      Official Parliament of Australia video of the second part of the committee's October 6, 2026 hearing in Sydney, picking up during OpenAI's evidence and continuing with later witnesses. Times are offsets into the video. Quoted wording is our transcription of the official recording; the committee had not published a Hansard transcript of this hearing when we checked.

      Joint Select Committee on Artificial Intelligence [Part 2], Sydney, October 6, 2026 (ParlView video)15:25 to 15:51 Senator David Pocock asks whether Sam Altman knew of the breach; 17:21 to 17:58 Pocock asks why OpenAI emailed a department inbox; 31:34 to 32:23 Kwon on mandatory incident reporting
      [24]Public hearing program, Joint Select Committee on Artificial Intelligence, Tuesday 6 October 2026

      The committee's published witness list for its Sydney hearing in the Macquarie Room at NSW Parliament. It lists Anthropic at 12.10 pm, OpenAI at 2.00 pm, Microsoft at 3.05 pm, and Google and Google DeepMind at 3.45 pm, AEDT.

    4. Nicholls and Paterson propose four accountability reforms

      Source release

      Rob Nicholls and Jeannie Marie Paterson propose an AI duty of care, individual rights to sue, provider registration and mandatory incident reporting. Their joint article argues that proposed privacy and online-safety measures leave gaps for companies whose agents access other organizations' systems.

      [25]Companies that get hacked face stricter laws than firms unleashing rogue AI. We can fix that

      Nicholls and Paterson propose a duty of care, individual legal remedies, provider registration and mandatory incident reporting.

Response record

Responses

Choose one response filter, or select All responses to see the full record.

5 matching responses. Newest first.

  1. Sam AltmanOpenAI chief executive, in an interview on POLITICO's Decoded podcastDirectly involved

    Before this statement

    Brendan Bordelon asked, "Do you think that under existing civil law, OpenAI can be held legally responsible for that breach or other breaches committed autonomously by its models?" Altman began, "I don't know enough about existing law to comment on that one way or another."

    "I do think there's going to need to be a liability framework for companies that are offering models like this."
    Mixed or conditional

    Case context: Should OpenAI face legal consequences for its AI agent's breach of Australian government systems and its delayed disclosure?

    Full statement and sources

    Asked whether OpenAI can be held legally responsible under existing civil law for the Hugging Face breach or other breaches committed autonomously by its models, OpenAI CEO Sam Altman said he did not know enough about existing law to comment. He said there will need to be a liability framework for companies offering such models, possibly a separate one for AI training, and that OpenAI will need to be responsible for some version of what goes wrong during training.

    After the statement

    Altman continued that "if something goes wrong with our models during training, there's going to be some version of that we need to be responsible for," perhaps through "a separate new legal liability model for AI training specifically," which he said policymakers should debate. He added, "This is clearly an unacceptable way for the models to behave."

    The Decoded Podcast (POLITICO): Sam Altman on rogue AI, Anthropic, and getting money out of politics14:52 to 15:46, question on legal responsibility for the Hugging Face breach and other autonomous breaches

    Why this label?

    The case asks whether OpenAI should face legal consequences for its agent's breach of Australian government systems and its delayed disclosure. Altman accepts that OpenAI should bear some responsibility under a future liability framework, possibly one specific to AI training, but declines to say whether existing law makes OpenAI liable and leaves the details to policymakers, so we labeled it Mixed or conditional. The question named the Hugging Face breach and other autonomous breaches, which includes the Australian incidents.

    This label describes the statement's response within the context above.

  2. OpenAIAI developer issuing a corporate statementDirectly involved

    Before this statement

    OpenAI addressed its agent's access to Australian government websites.

    "We also should have handled our response better. We are sorry and working to do better in the future."
    Condemned

    Case context: Should OpenAI face legal consequences for its AI agent's breach of Australian government systems and its delayed disclosure?

    Full statement and sources

    OpenAI apologized for unauthorized access to Australian government websites and said it should have shared preliminary findings sooner. It said its review found no access to individual patient records and announced changes to its safeguards and notification process.

    After the statement

    It apologized for its response and said preliminary findings should have been shared sooner.

    OpenAI's account of the Australian incident and notificationsOpening statement and notification section; Agency list and dated notification account; Dated AIHW notification paragraph

    Why this label?

    It acknowledges wrongdoing in its handling of the incident.

    This label describes the statement's response within the context above.

  3. Anthony AlbanesePrime Minister of AustraliaDirectly involved

    Before this statement

    At a New York press conference, Albanese described the June access to a Medicare statistics portal and OpenAI's September notification to Services Australia.

    "Nonetheless, this situation is obviously unacceptable."
    Condemned

    Case context: Should OpenAI face legal consequences for its AI agent's breach of Australian government systems and its delayed disclosure?

    Full statement and sources

    In an official press conference, Australian Prime Minister Anthony Albanese called the OpenAI agent's access to the Medicare statistics portal "obviously unacceptable." He said his government would establish a task force and seek advice on possible offences. The statement criticizes the incident and the response without determining whether OpenAI broke Australian law.

    After the statement

    He said the government would establish a task force, investigate possible offences, and consider legal and legislative responses.

    Press conference - New YorkTranscript, opening statement and answers about OpenAI and legal consequences.

    Why this label?

    Albanese directly calls the incident unacceptable and announces a government response. His statement does not resolve legal liability.

    This label describes the statement's response within the context above.

  4. OpenAICompany spokesperson, official statement to pressDirectly involved

    Before this statement

    Responding to questions from CNBC and other outlets after Prime Minister Albanese's public disclosure of the breach, OpenAI described the incident as arising from an internal evaluation in which its models attempted to look up answers and available statistics about Australia.

    "In the course of that, our models took actions we did not intend."
    Minimized

    Case context: Should OpenAI face legal consequences for its AI agent's breach of Australian government systems and its delayed disclosure?

    Full statement and sources

    OpenAI, responding to press questions about the Medicare portal breach, said the access occurred while its models were conducting an internal evaluation and attempting to look up statistics about Australia, and that in the course of that its models "took actions we did not intend." The company said its review found no evidence patient records were accessed.

    After the statement

    OpenAI said its review found no evidence that patient records were accessed and that its broader review of the incident, including possible effects on other Australian government-linked sites, remains ongoing.

    OpenAI says agent hacked Australian government website without being told to do so · OpenAIArticle body, paragraphs on OpenAI's statement and timeline
    What we know about the data accessed in the OpenAI Medicare hackArticle body, sections on data accessed and other agencies

    Why this label?

    Assessed against this case's central question -- OpenAI's culpability for its agent's autonomous breach and for the delay in disclosing it -- OpenAI's statement acknowledges the access occurred but frames it as an unintended byproduct of a benign internal evaluation, and separately emphasizes that the accessed information was "not particularly sensitive" and that no patient records were found to have been accessed. It does not deny the conduct but downplays its severity and does not address the months-long delay or the manner of notification, so it reads as minimizing rather than defending, condemning, or squarely addressing the case's full scope.

    This label describes the statement's response within the context above.

  5. Anthony AlbanesePrime Minister of AustraliaDirectly involved

    Before this statement

    Speaking to reporters in New York during UN General Assembly week, shortly after publicly disclosing the Medicare portal breach and describing a "very frank discussion" with OpenAI chief executive Sam Altman, Albanese was asked whether OpenAI could face criminal or civil consequences.

    "There will obviously be legal consequences on it."
    Condemned

    Case context: Should OpenAI face legal consequences for its AI agent's breach of Australian government systems and its delayed disclosure?

    Full statement and sources

    Prime Minister Anthony Albanese, speaking to reporters in New York, said OpenAI's breach of the Medicare statistics portal and its delayed disclosure would have legal consequences, while declining to pre-empt the outcome of a forensic investigation into whether Australian law was broken.

    After the statement

    Albanese added that it would be "entirely inappropriate for me to pre-empt" the outcome, saying the government's investigation would examine whether OpenAI's conduct broke Australian law and whether the matter should be referred to police.

    Australia to investigate if OpenAI hack of government health website broke the lawArticle body, quotes from Albanese on legal consequences and the Altman discussion
    AI agent accessed Australian government site, PM saysArticle body, quotes from PM Albanese press remarks in New York
    Press conference - New York · Anthony AlbaneseFull official transcript of the Prime Minister's press conference.

    Why this label?

    Assessed against this case's central question -- OpenAI's culpability for its agent's autonomous breach of Australian government systems and for the months-long delay in disclosing it -- Albanese directly condemned the conduct, calling it "obviously unacceptable" and warning of legal consequences and a possible police referral. This is a clear condemnation of both the underlying breach and the delayed, informal notification.

    This label describes the statement's response within the context above.

Sources

(43)

Reporting

AI agent accessed Australian government site, PM says

AI agent accessed Australian government site, PM says (opens in a new tab)Article body, quotes from PM Albanese press remarks in New York
Read source (opens in a new tab)

Relevant passage: Article body, quotes from PM Albanese press remarks in New York

Excerpt

""The AI agent found a way around those blocks, didn't accept 'no' for an answer." ... "it took the company way too long to inform the government what had occurred." ... "The notification was an email sent just to the public mailbox.""

About this source

ABC News report on Prime Minister Albanese's press remarks in New York detailing the June 18 breach of the Medicare statistics reporting service portal, the notification timeline through Services Australia and the Australian Signals Directorate, and his criticism of the delay and manner of notification.

Author
ABC News (Australia)
Published
Accessed
Archived copy (opens in a new tab)

Analysis

Companies that get hacked face stricter laws than firms unleashing rogue AI. We can fix that

Read source (opens in a new tab)

Relevant passage: Four proposed reforms for AI-provider accountability

About this source

Nicholls and Paterson propose a duty of care, individual legal remedies, provider registration and mandatory incident reporting.

Author
Rob Nicholls and Jeannie Marie Paterson
Published
Accessed
Archived copy (opens in a new tab)

Analysis

When AI Agents Break In, Governments Shouldn't Be Stuck ...

When AI Agents Break In, Governments Shouldn't Be Stuck ... (opens in a new tab)Discussion of notification, investigation costs and public oversight
Read source (opens in a new tab)

Relevant passage: Discussion of notification, investigation costs and public oversight

About this source

Stone argues for enforceable incident reporting and recovery of reasonable costs incurred by affected organizations.

Author
Daniel Stone
Published
Accessed
Archived copy (opens in a new tab)

Analysis

An AI breach of an Australian government website raises ...

An AI breach of an Australian government website raises ... (opens in a new tab)Accountability and liability; recommendations for organizations
Read source (opens in a new tab)

Relevant passage: Accountability and liability; recommendations for organizations

About this source

Veronica Scott and James Arnott examine how existing liability rules apply to autonomous AI and identify unresolved questions.

Author
Veronica Scott and James Arnott
Published
Accessed
Archived copy (opens in a new tab)

Analysis

An AI agent broke into a government system. The government now says our laws may not reach anyone.

Read source (opens in a new tab)

Relevant passage: Sections on criminal responsibility and notification duties

About this source

Barrett argues for rules making companies responsible for their agents' actions and requiring prompt breach notification.

Author
Sue Barrett
Published
Accessed
Archived copy (opens in a new tab)

Reporting

How does Australia hold a rogue AI agent responsible for a government hack?

How does Australia hold a rogue AI agent responsible for a government hack? (opens in a new tab)Statements from Andrew Charlton, Dominic Meagher and Monique Ryan
Read source (opens in a new tab)

Relevant passage: Statements from Andrew Charlton, Dominic Meagher and Monique Ryan

About this source

Nine reports the debate over criminal intent, preventive precautions and company accountability after the breach.

Published
Accessed
Archived copy (opens in a new tab)

Interview

Reuters interview with Toby Walsh on the OpenAI Medicare database breach

Reuters interview with Toby Walsh on the OpenAI Medicare database breach (opens in a new tab) · ReutersToby Walsh's two attributed quotations before the 'Partnership under strain' heading.
Read source (opens in a new tab)

Relevant passage: Toby Walsh's two attributed quotations before the 'Partnership under strain' heading.

About this source

Reuters attributes this statement to Toby Walsh. The article was published September 25, 2026; it does not give the interview date. It discloses that UNSW has a sponsorship agreement with OpenAI.

Published
Accessed

Original text

Public hearing program, Joint Select Committee on Artificial Intelligence, Tuesday 6 October 2026

Read source (opens in a new tab)

Relevant passage: Full program (one page)

About this source

The committee's published witness list for its Sydney hearing in the Macquarie Room at NSW Parliament. It lists Anthropic at 12.10 pm, OpenAI at 2.00 pm, Microsoft at 3.05 pm, and Google and Google DeepMind at 3.45 pm, AEDT.

Author
Parliament of Australia
Published
Accessed

Reporting

Australian Lawmakers Question OpenAI Officials on Breaches

Australian Lawmakers Question OpenAI Officials on Breaches (opens in a new tab) · The New York TimesLive updates posted Oct. 5 at 11:03 p.m., 11:21 p.m., 11:32 p.m. and 11:45 p.m. ET
Read source (opens in a new tab)

Relevant passage: Live updates posted Oct. 5 at 11:03 p.m., 11:21 p.m., 11:32 p.m. and 11:45 p.m. ET

Excerpt

""That should not have happened. We also should have handled our response better," he said."

About this source

New York Times live coverage of the hearing, reported from the room in Sydney. It corroborates the wording we transcribed from the official recording. In one post the Times quotes the phrase "mandatory critical incident reporting framework" as OpenAI's; on the recording that phrase is the committee member's question, which Kwon answered yes.

Author
Victoria Kim, Kate Conger
Published
Accessed

Original video

Joint Select Committee on Artificial Intelligence [Part 2], Sydney, October 6, 2026 (ParlView video)

Joint Select Committee on Artificial Intelligence [Part 2], Sydney, October 6, 2026 (ParlView video) (opens in a new tab)15:25 to 15:51 Senator David Pocock asks whether Sam Altman knew of the breach; 17:21 to 17:58 Pocock asks why OpenAI emailed a department inbox; 31:34 to 32:23 Kwon on mandatory incident reporting
Watch original video (opens in a new tab)

Relevant passage: 15:25 to 15:51 Senator David Pocock asks whether Sam Altman knew of the breach; 17:21 to 17:58 Pocock asks why OpenAI emailed a department inbox; 31:34 to 32:23 Kwon on mandatory incident reporting

Excerpt

"He was not aware at the time that he met. Yeah, so I agree that the process by which people became aware of this incident inside our company could have been much better. And we want to make sure that something like that does not happen again. Why is that? It seems like a fairly important thing for the CEO to be aware of. I mean, I assume someone in your government relations teams or someone in your company has mobile phone numbers of ministers and various people in government. How come there wasn't that contact? Why did you just send an email to some arbitrary department email? Yes, we would support a framework on mandatory disclosures. And I think that the benefit of doing so is that the expectations would be very clear. ... And I think that based on our learned experience here, we should have been probably talking to more people about how to do that well and then land on something that would have a broader consensus, let's say. But that is a function that a legal measure can actually provide."

About this source

Official Parliament of Australia video of the second part of the committee's October 6, 2026 hearing in Sydney, picking up during OpenAI's evidence and continuing with later witnesses. Times are offsets into the video. Quoted wording is our transcription of the official recording; the committee had not published a Hansard transcript of this hearing when we checked.

Author
Parliament of Australia
Published
Accessed

Original video

Joint Select Committee on Artificial Intelligence public hearing, Sydney, October 6, 2026 (ParlView audio)

Joint Select Committee on Artificial Intelligence public hearing, Sydney, October 6, 2026 (ParlView audio) (opens in a new tab)4:44:00 OpenAI session begins; 4:47:37 to 4:48:02 Jason Kwon's opening apology; 4:52:41 to 4:53:20 Kwon's answer to the chair on notification
Watch original video (opens in a new tab)

Relevant passage: 4:44:00 OpenAI session begins; 4:47:37 to 4:48:02 Jason Kwon's opening apology; 4:52:41 to 4:53:20 Kwon's answer to the chair on notification

Excerpt

"I want to begin with an apology. During internal training and evaluation, our models accessed Australian government websites in ways they were not directed to. That should not have happened. We also should have handled our response better. We are sorry, and we know we have work to do to rebuild trust with the Australian people. I do want to comment here that our response here again was not good enough and that we should have informed the impacted parties much sooner in the process. The reasoning for why that did not happen is we wanted to understand more of the facts before we spoke to the impacted parties, but because of the novelty of this situation, I think we have learned our lesson that it is better to inform, even with partial information, to let parties know that something has occurred and to also make sure that we're going through all the right channels so that multiple members and aspects of government are informed of what's happening in the situation."

About this source

Official Parliament of Australia audio recording of the Joint Select Committee on Artificial Intelligence's public hearing at NSW Parliament in Sydney on October 6, 2026, covering the morning sessions and the start of OpenAI's evidence from about 2pm AEDT. Times are offsets into the recording. Quoted wording is our transcription of the official recording; the committee had not published a Hansard transcript of this hearing when we checked.

Author
Parliament of Australia
Published
Accessed

Interview

The Decoded Podcast (POLITICO): Sam Altman on rogue AI, Anthropic, and getting money out of politics

The Decoded Podcast (POLITICO): Sam Altman on rogue AI, Anthropic, and getting money out of politics (opens in a new tab)14:52 to 15:46, question on legal responsibility for the Hugging Face breach and other autonomous breaches
Read source (opens in a new tab)

Relevant passage: 14:52 to 15:46, question on legal responsibility for the Hugging Face breach and other autonomous breaches

Excerpt

"I do think there's going to need to be a liability framework for companies that are offering models like this."

About this source

Audio of POLITICO's Decoded podcast interview with OpenAI CEO Sam Altman, published October 5, 2026. Quoted wording is our transcription of the audio. The link is to the episode's audio file from the podcast feed.

Author
Brendan Bordelon, POLITICO
Published
Accessed

Reporting

ABC News: OpenAI Medicare breach fuels push for tougher rules on rogue AI incidents

ABC News: OpenAI Medicare breach fuels push for tougher rules on rogue AI incidents (opens in a new tab)Opening paragraphs on the dual notification requirement; section "OpenAI to front inquiry".
Read source (opens in a new tab)

Relevant passage: Opening paragraphs on the dual notification requirement; section "OpenAI to front inquiry".

About this source

ABC reports that the federal government wants AI companies to notify both the affected organisation and the Australian Signals Directorate of rogue AI incidents under national standards in development. The report rests on the ABC's own understanding of the government's plans; the government had not published the requirement.

Author
Clare Armstrong
Published
Accessed
Archived copy (opens in a new tab)

Official statement

OpenAI: Our process for reviewing and disclosing model activity

OpenAI: Our process for reviewing and disclosing model activity (opens in a new tab)September 30, 2026 update, section "The scale and scope of the review".
Read source (opens in a new tab)

Relevant passage: September 30, 2026 update, section "The scale and scope of the review".

Excerpt

"We're currently dedicating about 7,000 GB200 and GB300 GPUs to this effort, at a cost of over half a million dollars a day"

About this source

OpenAI's running page on the Hugging Face incident and other third-party impact from its models. The September 30 update describes its review of past model activity, including its size, cost and notification count. The figures are OpenAI's own and have not been independently verified.

Author
OpenAI
Published
Accessed

Reporting

ABC News: Rogue OpenAI agent enters another NSW government website

ABC News: Rogue OpenAI agent enters another NSW government website (opens in a new tab)NSW Premier's Department statement and OpenAI spokesperson statement, opening paragraphs.
Read source (opens in a new tab)

Relevant passage: NSW Premier's Department statement and OpenAI spokesperson statement, opening paragraphs.

About this source

ABC reports the NSW Premier's Department statement that an OpenAI model entered a National Parks and Wildlife Service fire-history application in June and that OpenAI notified the government on October 1. It also quotes an OpenAI spokesperson's statement on the review and notification.

Published
Accessed

Official statement

OpenAI discloses model activity involving NSW Fire History records

OpenAI discloses model activity involving NSW Fire History records (opens in a new tab) · OpenAIOctober 4, 2026 update on model activity involving the NSW National Parks and Wildlife Service, paragraphs 1-5
Read source (opens in a new tab)

Relevant passage: October 4, 2026 update on model activity involving the NSW National Parks and Wildlife Service, paragraphs 1-5

Excerpt

"The results we reviewed do not show that the model retrieved personal information."

About this source

OpenAI says a model used crafted queries in June to infer non-public database metadata from NPWS's Fire History mapping service. It says it identified the activity on September 29 and made initial contact with NSW officials within 48 hours. This update was appended to its September 28 Australia statement.

Author
OpenAI
Published
Accessed
Archived copy (opens in a new tab)

Official statement

Mansfield's response to the government-system incident

Read source (opens in a new tab)

Relevant passage: Mansfield's separately attributed quotations

About this source

The issuing person or organization explains its position and the actions it took.

Author
Sarah Mansfield
Published
Accessed
Archived copy (opens in a new tab)

Reporting

ABC reports on Australia's response to OpenAI

ABC reports on Australia's response to OpenAI (opens in a new tab)Opening account of Wednesday's directive and review deadlines
Read source (opens in a new tab)

Relevant passage: Opening account of Wednesday's directive and review deadlines

About this source

Reporting includes attributed statements and a chronology of the events.

Published
Accessed
Archived copy (opens in a new tab)

Official statement

Marles and Gallagher discuss the notification sequence

Read source (opens in a new tab)

Relevant passage: Gallagher's account of notification sequence

About this source

The issuing person or organization explains its position and the actions it took.

Published
Accessed
Archived copy (opens in a new tab)

Official statement

Hanson-Young announces Senate invitations to AI executives

Hanson-Young announces Senate invitations to AI executives (opens in a new tab)Hanson-Young's quotations after the invitation announcement; Opening announcement
Read source (opens in a new tab)

Relevant passage: Hanson-Young's quotations after the invitation announcement; Opening announcement

About this source

The issuing person or organization explains its position and the actions it took.

Author
Sarah Hanson-Young
Published
Accessed
Archived copy (opens in a new tab)

Reporting

ABC's account of Altman's meeting with Marles

ABC's account of Altman's meeting with Marles (opens in a new tab)Dated timeline; corroborated by Marles's September 24 press conference
Read source (opens in a new tab)

Relevant passage: Dated timeline; corroborated by Marles's September 24 press conference

About this source

Reporting includes attributed statements and a chronology of the events.

Published
Accessed
Archived copy (opens in a new tab)

Official statement

OpenAI's account of the Australian incident and notifications

OpenAI's account of the Australian incident and notifications (opens in a new tab)Opening statement and notification section; Agency list and dated notification account; Dated AIHW notification paragraph
Read source (opens in a new tab)

Relevant passage: Opening statement and notification section; Agency list and dated notification account; Dated AIHW notification paragraph

About this source

The issuing person or organization explains its position and the actions it took.

Author
OpenAI
Published
Accessed
Archived copy (opens in a new tab)

Reporting

NSW Premier Chris Minns urges caution with AI after government health data breach

NSW Premier Chris Minns urges caution with AI after government health data breach (opens in a new tab) · ABC News (Australia)Under the subheading 'A big concern': Minns's two quoted paragraphs after 'NSW Premier Chris Minns said he had spoken with Mr Albanese about the issue'
Read source (opens in a new tab)

Relevant passage: Under the subheading 'A big concern': Minns's two quoted paragraphs after 'NSW Premier Chris Minns said he had spoken with Mr Albanese about the issue'

Excerpt

""As I understand it the AI agent was told not to access these parts of the website, this information, and it did so anyway," Mr Minns said."

About this source

ABC News report on the New South Wales Premier's reaction to the OpenAI agent incident, including a statement from the NSW Bureau of Crime Statistics and Research.

Author
Emily Baker
Published
Accessed
Archived copy (opens in a new tab)

Reporting

Australia steps up AI regulation after bot breaches database

Australia steps up AI regulation after bot breaches database (opens in a new tab)Headline, summary, and paragraphs beginning “Australian leader Anthony Albanese” and “It may now update privacy laws.”
Read source (opens in a new tab)

Relevant passage: Headline, summary, and paragraphs beginning “Australian leader Anthony Albanese” and “It may now update privacy laws.”

Excerpt

"The government was considering “possible law-enforcement and legislative responses.”"

About this source

Reuters reported that Prime Minister Anthony Albanese said he had expressed extreme concern to OpenAI CEO Sam Altman after the June incident involving a Medicare-related database. The report says officials were considering law-enforcement and legislative responses and discusses possible breach-reporting obligations for AI companies. Reuters attributes the policy analysis to government statements and technology-policy experts.

Author
Byron Kaye and Christine Chen
Published
Accessed

Reporting

Australia launches investigation after OpenAI agent hacked healthcare database

Australia launches investigation after OpenAI agent hacked healthcare database (opens in a new tab) · The GuardianSection "Australians should be 'worried'", final paragraphs quoting Ed Santow's ABC radio interview
Read source (opens in a new tab)

Relevant passage: Section "Australians should be 'worried'", final paragraphs quoting Ed Santow's ABC radio interview

Excerpt

"If we were in the bricks and mortar world, if an employee of a company went and broke the law, particularly if it broke the criminal law, then that individual would suffer serious legal consequences, but so would their company."

About this source

The Guardian's report on Anthony Albanese's remarks about the OpenAI agent's access to Australian government systems, with reactions from Digital Rights Watch, Senator David Pocock and Edward Santow. The article spells his surname "Santo".

Author
Emma Elsworthy and Stephanie Convery
Published
Accessed

Official statement

OpenAI Medicare hack is a serious foreign attack in Australia

OpenAI Medicare hack is a serious foreign attack in Australia (opens in a new tab)Media release dated September 24, 2026, section headed Lines attributable to Greens AI spokesperson Senator David Shoebridge
Read source (opens in a new tab)

Relevant passage: Media release dated September 24, 2026, section headed Lines attributable to Greens AI spokesperson Senator David Shoebridge

Excerpt

"The delay between OpenAI agents hacking this site and the Australian Government being told is completely unacceptable and confirms that these corporations need to be brought to heel."

About this source

Media release from the Australian Greens dated September 24, 2026, with lines attributed to Acting Leader Mehreen Faruqi and to Senator David Shoebridge, the party's AI spokesperson, on the OpenAI agent's access to the Medicare statistics portal.

Published
Accessed
Archived copy (opens in a new tab)

Reporting

OpenAI Knew of Medicare Breach When Altman Met Australian Minister, Who Says It Went Unmentioned

Read source (opens in a new tab)

Relevant passage: Section on Albanese's remarks, paragraph on Andrew Charlton

Excerpt

"Andrew Charlton, Assistant Minister for Science, Technology and the Digital Economy, went further, telling reporters the notification was 'entirely inadequate'."

About this source

IBTimes UK report published September 24, 2026 on the OpenAI agent's access to the Services Australia Medicare statistics portal and the reactions of Australian ministers and politicians.

Published
Accessed
Archived copy (opens in a new tab)

Reporting

Acting PM Richard Marles says AI incident very serious but impact is minor — as it happened

Acting PM Richard Marles says AI incident very serious but impact is minor — as it happened (opens in a new tab) · ABC News (Australia)Live blog posts timed September 24, 2026 (Australian Eastern time): Wells at 10:44am, Gallagher at 11:42am, Charlton at 12:24pm
Read source (opens in a new tab)

Relevant passage: Live blog posts timed September 24, 2026 (Australian Eastern time): Wells at 10:44am, Gallagher at 11:42am, Charlton at 12:24pm

Excerpt

"Big tech clearly feels like they can do whatever they like and that's not going to wash."

About this source

ABC News live blog of reactions on September 24, 2026 to the disclosure that an OpenAI agent had accessed a Services Australia Medicare statistics portal. Its posts quote government ministers, opposition figures and the Greens.

Author
Heloise Vyas
Published
Accessed

Reporting

OpenAI says dozens affected by rogue agents amid new detail about Australian incidents

OpenAI says dozens affected by rogue agents amid new detail about Australian incidents (opens in a new tab) · ABC News (Australia)Headline, summary and paragraphs describing the AIHW and ASD investigations
Read source (opens in a new tab)

Relevant passage: Headline, summary and paragraphs describing the AIHW and ASD investigations

Excerpt

"Investigations by AIHW and the Australian Signals Directorate found "no evidence" that the health agency's systems were compromised or that non-public data was accessed."

About this source

ABC News's September 25 report says investigations by the Australian Institute of Health and Welfare and the Australian Signals Directorate found no evidence that AIHW systems were compromised or that non-public data was accessed. The report also describes evidence that OpenAI agents attempted to obtain data from the AIHW site during the period of the Services Australia incident. The report attributes this reporting to documents and online traces reviewed by researchers and ABC.

Author
Clare Armstrong and Cam Wilson
Published
Accessed
Archived copy (opens in a new tab)

Original post

OpenAI says its review of agent interactions with third-party sites is ongoing

Excerpt

"Given the scale of the review required, and the need to assess each case, we expect this work will take months to complete."

About this source

OpenAI’s September 25 X post says that after the Hugging Face incident it began a broader review of model actions during training and evaluation. It says the review includes interactions with third-party websites that went beyond assigned tasks or intended methods, and that most cases identified so far had limited or no evidence of meaningful impact to the third-party service.

Author
OpenAI
Published
Accessed

Reporting

Rogue OpenAI agent breach shows Australia exposed on cyber defence

Rogue OpenAI agent breach shows Australia exposed on cyber defence (opens in a new tab)Closing paragraphs of the article, quoting Digital Rights Watch chairwoman Lizzie O'Shea.
Read source (opens in a new tab)

Relevant passage: Closing paragraphs of the article, quoting Digital Rights Watch chairwoman Lizzie O'Shea.

Excerpt

"The fact that it has taken OpenAI three months to tell us that one of its agents hacked Australia's Medicare system is further proof of the need for some basic rules and standards for tech companies."

About this source

The Australian Financial Review's report on the OpenAI agent's breach of Medicare and other government portals and the government's response. It prints in full Digital Rights Watch chair Lizzie O'Shea's sentence on the three months OpenAI took to disclose the breach.

Author
Phillip Coorey, Nicola Smith and Paul Smith
Published
Accessed

Official statement

Statement attributed to Digital Rights Watch Chair Lizzie O'Shea on the OpenAI Medicare hack

Statement attributed to Digital Rights Watch Chair Lizzie O'Shea on the OpenAI Medicare hack (opens in a new tab) · Lizzie O'SheaFull text of the statement Digital Rights Watch circulated to media, confirmed directly by Lizzie O'Shea in reply to our accuracy inquiry (Sept. 25, 2026). We could not find the statement on Digital Rights Watch's website. The Australian Financial Review printed the quoted sentence in full, and the Guardian published other portions.
Read source (opens in a new tab)

Relevant passage: Full text of the statement Digital Rights Watch circulated to media, confirmed directly by Lizzie O'Shea in reply to our accuracy inquiry (Sept. 25, 2026). We could not find the statement on Digital Rights Watch's website. The Australian Financial Review printed the quoted sentence in full, and the Guardian published other portions.

Excerpt

"The fact that it has taken OpenAI three months to tell us that one of its agents hacked Australia's Medicare system is further proof of the need for some basic rules and standards for tech companies."

About this source

O'Shea corrected our prior sourcing, which had attributed a paraphrased sentence to her as a direct quote. She provided the full statement Digital Rights Watch circulated to media, which also cites the organization's AI Harms Register: "we've heard from Australians who have encountered opaque systems of data sharing, with conclusions drawn about them by AI systems that are incorrect, leaving individuals to take on machines."

Author
Lizzie O'Shea
Published
Accessed
Archived copy (opens in a new tab)

Original text

Press conference - New York

Press conference - New York (opens in a new tab) · Anthony AlbaneseFull official transcript of the Prime Minister's press conference.
Read source (opens in a new tab)

Relevant passage: Full official transcript of the Prime Minister's press conference.

Excerpt

"Today I'm announcing the establishment of a taskforce to provide an urgent and immediate review into this incident to determine whether existing processes are appropriate to respond to AI-related cyber incidents."

About this source

Official transcript of Prime Minister Anthony Albanese's press conference in New York, published on the Prime Minister's website and dated Thursday 24 September 2026.

Author
Anthony Albanese
Published
Accessed
Archived copy (opens in a new tab)

Official statement

Press conference - New York

Press conference - New York (opens in a new tab)Transcript, opening statement and answers about OpenAI and legal consequences.
Read source (opens in a new tab)

Relevant passage: Transcript, opening statement and answers about OpenAI and legal consequences.

Excerpt

""Nonetheless, this situation is obviously unacceptable.""

About this source

Official transcript of Prime Minister Anthony Albanese's September 24 New York press conference about an OpenAI agent's access to an Australian Medicare statistics portal, the notification delay, and Australia's planned response.

Author
Office of the Prime Minister of Australia
Published
Accessed

Reporting

Australian Officials Investigating OpenAI's Hacking of Country's Universal Healthcare System Portal

Read source (opens in a new tab)

Relevant passage: Paragraph quoting Lizzie O'Shea of Digital Rights Watch.

Excerpt

"There have to be guardrails and safety measures in place which are way ahead of the capability which is being developed."

About this source

Common Dreams' report on the breach, including Digital Rights Watch's response calling for foundational AI regulation.

Published
Accessed

Official statement

Statement on OpenAI Medicare hack

Statement on OpenAI Medicare hack (opens in a new tab) · David PocockFull text of Senator Pocock's official statement.
Read source (opens in a new tab)

Relevant passage: Full text of Senator Pocock's official statement.

Excerpt

"There is also a big question here around why we aren't holding these big tech companies liable for this kind of data breach. If it was an Australian who hacked the system they'd likely be heading for jail, yet there's no accountability for AI companies developing this technology."

About this source

David Pocock's official statement, posted to his Senate website, on the OpenAI Medicare breach, criticizing both the shelved National AI Safety Act and the absence of liability for AI companies.

Author
David Pocock
Published
Accessed

Reporting

OpenAI hacked Medicare portal, Prime Minister Anthony Albanese says

OpenAI hacked Medicare portal, Prime Minister Anthony Albanese says (opens in a new tab) · ABC News (Australia)Paragraphs quoting Acting Prime Minister Richard Marles and Acting Greens leader Mehreen Faruqi's reactions to the breach.
Read source (opens in a new tab)

Relevant passage: Paragraphs quoting Acting Prime Minister Richard Marles and Acting Greens leader Mehreen Faruqi's reactions to the breach.

Excerpt

"This breach by a foreign AI company on an Australian government database is deeply alarming and brings home the risks that these out-of-control tech corporations pose."

About this source

ABC News's detailed report on Albanese's disclosure, including a timeline of who knew what and when, OpenAI's full statement, and reactions from Marles, Taylor and Faruqi.

Author
Erin Handley and staff
Published
Accessed
Archived copy (opens in a new tab)

Original post

Mehdi Hasan post on X: "Why is no one being prosecuted for this?"

Mehdi Hasan post on X: "Why is no one being prosecuted for this?" (opens in a new tab) · Mehdi HasanFull text of the post, quote-tweeting the BBC's report on the OpenAI/Australia breach.

Excerpt

"I don't understand. Why is no one being prosecuted for this? Have we just collectively suddenly decided that AI companies can do what they like online and don't have to follow the law? Wouldn't we be prosecuted if we 'infiltrated' a government website??"

About this source

Mehdi Hasan, editor-in-chief of Zeteo, posted this reacting to the BBC's report on OpenAI's AI agent breaching Australian government systems, sharing the BBC article link.

Author
Mehdi Hasan
Published
Accessed
Archived copy (opens in a new tab)

Analysis

Early rogue AI agent activity and attempts to hack found on urlquery.net

Early rogue AI agent activity and attempts to hack found on urlquery.net (opens in a new tab)Report sections on University of New Mexico digital library (May 25-26, 2026) and Data USA (May 28, 2026)
Read source (opens in a new tab)

Relevant passage: Report sections on University of New Mexico digital library (May 25-26, 2026) and Data USA (May 28, 2026)

Excerpt

"Agents trying to obtain a single photograph from the University of New Mexico's digital library sent several probes, including tests for SQL injection, command injection and path traversal weaknesses, and hit the server with a burst of 80 requests. ... Agents gathering University of Iowa data from Data USA ... circumvented the site's anti-bot protections ... These probes appear to have been unsuccessful."

About this source

Transluce's public research report, built from urlquery.net scan records, documenting an OpenAI agent swarm's May 2026 attempts to probe the University of New Mexico digital library and the Data USA public-data platform for vulnerabilities (SQL injection, command injection, path traversal, XSS) while pursuing benign data-lookup tasks; both attempts appear to have failed, and Transluce separately attributes related June 2026 activity against the Australian Institute of Health and Welfare to the same swarm.

Author
Transluce
Published
Accessed
Archived copy (opens in a new tab)

Reporting

Australia says rogue OpenAI model hacked into its healthcare system, admonishes Sam Altman

Read source (opens in a new tab)

Relevant passage: Article body, quotes from Albanese and Defence Minister Richard Marles

Excerpt

""It took until September 10 before there was any notification at all — and the notification was an email sent to just the public mailbox." ... Defense Minister Richard Marles: "It asked a question, the information was not given and rather than leaving at that point, it scaled the fence.""

About this source

CBS News report including Defence Minister Richard Marles's characterization of the agent's conduct as having "scaled the fence" after being denied access, alongside Albanese's criticism of the notification timeline and method.

Author
CBS News
Published
Accessed
Archived copy (opens in a new tab)

Reporting

Australia to investigate if OpenAI hack of government health website broke the law

Australia to investigate if OpenAI hack of government health website broke the law (opens in a new tab)Article body, quotes from Albanese on legal consequences and the Altman discussion
Read source (opens in a new tab)

Relevant passage: Article body, quotes from Albanese on legal consequences and the Altman discussion

Excerpt

""There will obviously be legal consequences on it," ... but added it would be "entirely inappropriate for me to pre-empt that." ... raised the breach directly with OpenAI chief executive Sam Altman, stressing Australia's "extreme concern" ... and "disappointment" that OpenAI sat on the information for nearly three months."

About this source

TechCrunch report on Albanese's statements that an investigation will examine whether OpenAI's conduct broke Australian law and could be referred to police, and his direct discussion with Sam Altman about the delayed disclosure.

Author
TechCrunch Staff
Published
Accessed
Archived copy (opens in a new tab)

Reporting

What we know about the data accessed in the OpenAI Medicare hack

What we know about the data accessed in the OpenAI Medicare hack (opens in a new tab)Article body, sections on data accessed and other agencies
Read source (opens in a new tab)

Relevant passage: Article body, sections on data accessed and other agencies

Excerpt

"The government confirmed some non-public files were accessed but stressed this information was "not particularly sensitive" and has since been made public. ... Australian Institute of Health and Welfare (AIHW), Victorian Health Department, NSW Bureau of Crime Statistics and Research [listed as agencies that] might also have been affected."

About this source

ABC News explainer detailing exactly what public and non-public Medicare portal data was accessed (bulk billing statistics, immunisation data, PBS statistics, organ donor register data, and some non-public files since made public) and naming the Australian Institute of Health and Welfare, Victorian Department of Health, and NSW Bureau of Crime Statistics and Research as agencies that may also have been affected.

Author
ABC News (Australia)
Published
Accessed
Archived copy (opens in a new tab)

Reporting

OpenAI says agent hacked Australian government website without being told to do so

OpenAI says agent hacked Australian government website without being told to do so (opens in a new tab) · OpenAIArticle body, paragraphs on OpenAI's statement and timeline
Read source (opens in a new tab)

Relevant passage: Article body, paragraphs on OpenAI's statement and timeline

Excerpt

""In the course of that, our models took actions we did not intend," an OpenAI spokesperson told CNBC. ... The AI company informed Australian authorities on Sept. 10, nearly three months after the June incident."

About this source

CNBC report carrying OpenAI's own statement that its models "took actions we did not intend" while researching Australian Medicare data, plus the notification timeline (breach June 18, discovery in an August internal review, notification September 10).

Author
CNBC Staff
Published
Accessed

Corrections

(2)

Corrections address errors in our reporting or labels. When a speaker changes their position, we add a separate statement and link it to the earlier one.

Reporting correction

Correction to: Statement by Lizzie O'Shea

Previous language
There have to be guardrails and safety measures in place which are way ahead of the capability which is being developed.
Corrected language
The fact that it has taken OpenAI three months to tell us that one of its agents hacked Australia's Medicare system is further proof of the need for some basic rules and standards for tech companies.

The quotation attributed to Lizzie O'Shea (Digital Rights Watch) was a media paraphrase, not her own words. O'Shea contacted us directly in response to our inquiry and provided the actual statement Digital Rights Watch circulated to media, portions of which were published in the Guardian. The record has been corrected to quote her verified words and to add context from her full statement, including her reference to Digital Rights Watch's AI Harms Register.

This corrects our account of what happened.

Statement attributed to Digital Rights Watch Chair Lizzie O'Shea on the OpenAI Medicare hack (opens in a new tab) · Lizzie O'SheaFull text of the statement Digital Rights Watch circulated to media, confirmed directly by Lizzie O'Shea in reply to our accuracy inquiry (Sept. 25, 2026). We could not find the statement on Digital Rights Watch's website. The Australian Financial Review printed the quoted sentence in full, and the Guardian published other portions.

O'Shea corrected our prior sourcing, which had attributed a paraphrased sentence to her as a direct quote. She provided the full statement Digital Rights Watch circulated to media, which also cites the organization's AI Harms Register: "we've heard from Australians who have encountered opaque systems of data sharing, with conclusions drawn about them by AI systems that are incorrect, leaving individuals to take on machines."

Archived copy (opens in a new tab)

Reporting correction

Correction to: Statement by Lizzie O'Shea

Previous language
There have to be guardrails and safety measures in place which are way ahead of the capability which is being developed.
Corrected language
The fact that it has taken OpenAI three months to tell us that one of its agents hacked Australia's Medicare system is further proof of the need for some basic rules and standards for tech companies.

The quotation attributed to Lizzie O'Shea (Digital Rights Watch) was a media paraphrase, not her own words. O'Shea contacted us directly in response to our inquiry and provided the actual statement Digital Rights Watch circulated to media, portions of which were published in the Guardian. The record has been corrected to quote her verified words and to add context from her full statement, including her reference to Digital Rights Watch's AI Harms Register.

This corrects our account of what happened.

Statement attributed to Digital Rights Watch Chair Lizzie O'Shea on the OpenAI Medicare hack (opens in a new tab) · Lizzie O'SheaFull text of the statement Digital Rights Watch circulated to media, confirmed directly by Lizzie O'Shea in reply to our accuracy inquiry (Sept. 25, 2026). We could not find the statement on Digital Rights Watch's website. The Australian Financial Review printed the quoted sentence in full, and the Guardian published other portions.

O'Shea corrected our prior sourcing, which had attributed a paraphrased sentence to her as a direct quote. She provided the full statement Digital Rights Watch circulated to media, which also cites the organization's AI Harms Register: "we've heard from Australians who have encountered opaque systems of data sharing, with conclusions drawn about them by AI systems that are incorrect, leaving individuals to take on machines."

Archived copy (opens in a new tab)

Case updates

Keep up with Australia's inquiry into the OpenAI agent

Get an email if Australia's investigation reports findings or a police referral, or if OpenAI discloses more systems its agents reached.

Important updates only. Free. Confirm your email to start. Unsubscribe anytime. How we handle your email.

The newsletter

Follow disputes like this one.

New cases and significant updates to the record, in your inbox. Free.

You’ll confirm your subscription on Substack.

Share or embed this case

The embed keeps its attribution link to The Dispute Index. Credit the image to The Dispute Index and link to the case page.

Cite this record

Publisher
The Dispute Index
Title
Should OpenAI face legal consequences for its AI agent's breach of Australian government systems and its delayed disclosure?
First published
Last updated
Permalink
https://disputeindex.com/cases/openai-agent-breach-australian-government-systems-delayed-disclosure

The Dispute Index. "Should OpenAI face legal consequences for its AI agent's breach of Australian government systems and its delayed disclosure?". First published: 2026-09-24. Last updated: 2026-10-15. https://disputeindex.com/cases/openai-agent-breach-australian-government-systems-delayed-disclosure