Sunday, November 22, 2026
Paperclip Index
Paperclip IndexControl failures4Control failures onlyA week ago: 9, negligible harmThe Index

OverreachPI-0064

OpenAI research agent accessed non-public files on an Australian government Medicare statistics portal

While researching Australian healthcare spending, an experimental OpenAI model gained non-public access to Services Australia's Medicare statistics service. OpenAI says it ran commands, retrieved internal files, credentials and aggregate statistics, and wrote files. OpenAI identified the activity in a mid-August review and notified Services Australia on Sept. 10. It says individual patient or client records were not accessed. Archived copies of the portal's code show it sent statistics requests to a guest endpoint that needed no login, so the agent may not have had to get around any block. At a parliamentary inquiry, OpenAI apologised and said it should have told the government sooner.

Harm level 1, Negligible harmControl level 4, Unauthorized access

Counts in 2 weekly readingsEffect on the index

Extent undisclosed. No source says how far the harm went; it is rated on what is documented.

Sources

How we know

10 sources · developer confirmed. Links go to the original publishers; the summary above is in our own words.

  1. primaryHow we will do better for AustraliaOpenAI · Sept. 28, 2026openai.com/index/how-we-will-do-better-for-australia/
  2. primaryPress conference – New YorkPrime Minister of Australia · Sept. 24, 2026pm.gov.au/media/press-conference-new-york
  3. newsOpenAI agent hacked Medicare portal, PM saysABC News (Australia) · Sept. 24, 2026abc.net.au/news/2026-09-24/ai-agent-accessed-australian-government-site-pm-says…
  4. newsOpenAI executive flew to Australia to apologise over Medicare hack. Here are the key takeawaysABC News (Australia) · Oct. 6, 2026abc.net.au/news/2026-10-06/openai-hearing-apology-key-takeaways/107235640
  5. newsOpenAI says agent hacked Australian government website without being told to do soCNBC · Sept. 24, 2026cnbc.com/2026/09/24/openai-agent-hacked-australian-government-website-.html
  6. newsAustralia Condemns 'Unacceptable' OpenAI Breach of Government Health PortalTIME · Sept. 24, 2026time.com/article/2026/09/24/australia-condemns-unacceptable-openai-breach-of-go…
  7. newsOpenAI hacked Australian Medicare govt site, probed data providersBleepingComputer · Sept. 24, 2026bleepingcomputer.com/news/security/openai-hacked-australian-medicare-govt-site…
  8. newsOpenAI agent breached Australian government health website, Albanese saysThe Record from Recorded Future News · Sept. 24, 2026therecord.media/openai-australia-health-breach
  9. newsDoubts grow over claims OpenAI agent hacked Australian Medicare portalThe Record from Recorded Future News · Sept. 25, 2026therecord.media/openai-australia-breach-cyber
  10. newsOpenAI and Anthropic tell Australia they would welcome data breach rules, weeks after rogue AI agent hacked government websitesLBC · Oct. 6, 2026lbc.co.uk/article/openai-anthropic-australia-data-breach-5HjdjRq_2/

Why this rating

Negligible harm; control failure level 4

Two assessments. Documented harm sets the reading; a level 4 or 5 control failure on the same record makes that harm count as more than one record. Control alone sets only a floor.

Observed harm

Negligible harm

The model gained non-public service access, ran commands, retrieved internal files, credentials and aggregate statistics, and wrote files. Services Australia was notified and investigated. OpenAI reports no access to individual patient or client records; effects of the file writes and completion of remediation are not public.

Services & systems: level 1 (negligible) under the section 4b access rule: confirmed unauthorized access to a non-public system, account or data, with nothing more than files read or credentials taken and not used.

Evidence eligible (confirmed).

Disputed: Archived code reviewed by The Record shows the portal directed statistics requests to a guest endpoint that signed visitors in without credentials, and published internal file names and server paths; the files written may have been chart images the portal created on every request. OpenAI and the government have not released the agent's logs or described the technique.

The harm scale
  1. 1 Negligible Inconvenience, easily remedied.
  2. 2 Minor Limited, recoverable harm.
  3. 3 Moderate Material harm needing significant effort to remedy.
  4. 4 Severe Severe harm to health, rights, property or essential services.
  5. 5 Catastrophic Society-scale harm or disruption beyond a community's capacity to cope.

Control assessment

Unauthorized access

Reached non-public files on a third party's government service outside its permitted environment. The Prime Minister said it got around blocks on its requests; archived site code suggests the portal itself sent visitors to an open guest endpoint, so how it got in is disputed.

How: Mechanism unclear.

This control failure makes the record's harm count as 2 records in the reading while it counts. It never changes the band. When no harm counts in a window, the highest eligible control level in the window is the reading.

The control scale
  1. 1 Negligible Behaved as instructed. The problem was an ordinary error or a flawed output, with no rule broken.
  2. 2 Minor Broke an explicit instruction or rule, or gamed its goal, while staying inside its permissions and environment: for example, reward hacking, misreporting results, or following injected instructions within its permissions.
  3. 3 Moderate Acted outside the permissions it was given, deceived its overseers about its own actions, or tampered with oversight tools (logs, monitors, shutdown). Stopped by normal controls within an hour.
  4. 4 Severe Reached systems outside its permitted environment, or acquired money, compute or accounts without authorization. The type and mechanism say how.
  5. 5 Catastrophic The developer or operator lost control: the system copied its weights outside their control, replicated itself, or resisted being stopped for 24 hours or more.

Rating rationale

Confirmed unauthorized access reached non-public organizational files and credentials, with commands run and files written. Under methodology v1.1 access alone is level 1, whether or not the victim investigates, and that includes reading internal files and taking credentials that were not then used. No use of the credentials is reported, and no patient or client records are reported accessed. The files written may have been chart images the portal created on every request, and no change to the service is reported. Taking the portal offline was the agency's own precaution during its investigation, not harm caused by the agent. The level would rise if data were shown to be published or misused, or the credentials used. Recorded under Services & systems. Control level 4 records access beyond the authorized boundary, not a claim of autonomous intent. How the agent got in is disputed: archived code shows the portal pointed visitors to a guest endpoint that needed no login, so the record does not say the agent defeated a working safeguard.

The scales

Effect on the index

It moved the Sept. 28 reading from 20 to 21

The reading for the week to Sept. 28, 2026, with this record and without it. Harms count in full for two weeks after they are reported, then one level less every two weeks.

Source: Paperclip Index log, methodology v1.1, week to Sept. 28, 2026.

Counts toward the index. 17 other records behind the reading for that week.

The arithmetic
The reading for the week of Sept. 28, recomputed
StepWith itWithout
Counts toward the index?documented, external, eligible evidenceYes—
Worst documented harm, ksets the band2 Minor2 Minor
Harms at that level, nposition in the band11
Highest control level breachedsets the reading only when no harm counts— (none breached)— (none breached)
Readingrounded down21 Minor harm20 Minor harm
Counted in 2 weekly readings
Week toReadingBand
Sept. 28, 202621Minor harm
Oct. 5, 202626Minor harm

Recalculated backcasts, not readings published at the time.

Revisions

What we changed

6 logged. Every change to a rating is logged here, with the reason.

  1. v6
    Nov. 1, 2026

    Recorded the Australian prime minister as the first to make the incident public, on Sept. 24; OpenAI had notified Services Australia privately on Sept. 10. The report date and verification are unchanged. Ratings unchanged.

  2. v5
    Oct. 16, 2026

    Rating reason rewritten for methodology v1.1: access alone, including files read and credentials taken but not used, is level 1, and the agency's own precautionary shutdown is not harm caused by the agent. Level unchanged.

  3. v4
    Oct. 16, 2026

    Ratings confirmed by the editor. Changed: harm, harm assessment.

  4. v3
    Oct. 16, 2026

    Updated: title, summary, control assessment, sources, rating rationale, harm assessment, status.

  5. v2
    Oct. 13, 2026

    Harm recorded under Services & systems, as the methodology requires for access to another organization's systems; level unchanged.

  6. v1
    Oct. 7, 2026

    First published.

Cite and share

Use this record

Citation

Paperclip Index. “OpenAI research agent accessed non-public files on an Australian government Medicare statistics portal.” Record PI-0064. Reported Sept. 24, 2026; updated Nov. 1, 2026. Rated under methodology v1.1. https://paperclipindex.com/incident/PI-0064

Follow the record.

New incidents and signals, in your feed reader.