The Automated Panopticon
Also published on FOI Warfare · Short reading added 15 November 2026. Earlier full version preserved below.
TRUTH SEQUENCES · SATIRE
The risk-scoring machine is fictional. The quoted words come from Services Australia’s FOI release LEX 93986.
REVIEW TRIGGER Customer literacy
The Customer reads the record. Worse, the Customer reads earlier parts of the record. Then the Customer compares dates.
Which email?
1 July 2025 · printed p. 21
A case officer records telling the Customer that he had threatened the officer in an email. The Customer, in the officer’s words, “asked which email”. The officer records advising him that Services Australia “does not need to provide evidence to justify or explain his restrictions”.
Under Behaviour and Warnings, the officer finds room for four words.
“Demanding, counter-productive, paranoid, Confused.”
Turn one page forward and six days back. On 25 June another entry says the Customer “denied having sent the threatening material” (printed p. 22). The Customer now has four adjectives. He still has a question about an email.
September gets its own page
15 September 2025 · printed p. 145 · PDF position 140
A separate incident entry quotes part of an email and records “Nil further action taken” because “no direct/active threat has been made by Customer”.
The machine would like one finding for the whole file. The Customer has put this one under September.
The third line
August 2026 review · printed p. 5
“Customer invited to participate in the review: Yes”
“Customer spoken to about the review: No”
Keep reading. The review says he “responded to the desktop SMS”. The form gets two boxes. His reply gets a bullet point.
Earlier full satire — preserved for context
The Automated Panopticon
What happens when a bureaucratic record meets a Customer who kept the paperwork?
Based on Services Australia release LEX 93986.
The review counts over 270 emails across the periods it lists.
It describes the threatening nature of some correspondence.
It alleges I signed rent certificates using aliases. Those are the agency’s claims. Read page 3 and page 4.
REVIEW TRIGGER: One file, two ways of numbering its pages
STATUS: PERMANENTLY INCONVENIENT
AI CONFIDENCE: Approaching existential panic
1. The intake problem
The automated review has found an anomaly for which no servicing code appears adequate.
The Customer reads the record. Worse, the Customer reads earlier parts of the record. Then the Customer compares dates.
When an allegation is recorded, the Customer asks for the evidence. One case note records:
“Customer demanded evidence of this, asked which email…”
The same case note records what the One Main Contact told the Customer:
“Services Australia does not need to provide evidence to justify or explain his restrictions, this is a decision the agency is able to make.” [Case note · p. 21]
Forensic AI observation: the Customer has made the elementary error of assuming that an allegation capable of producing a restriction might also be capable of producing its evidence. Circularity detected.
2. Behaviour and warnings
The record describes the Customer in a case note as:
“counter-productive, paranoid” [p. 21]
The August review mentions the volume of correspondence and the Customer’s use of AI when it recommends how the agency should handle future contact. [Review · p. 3] [Recommendations · p. 4]
The Customer responded by building a source index. It separates printed page numbers from PDF positions, so anyone can find the page being described. [Page register]
That creates the first major system vulnerability: PDF position ≠ printed page number.
Updated behavioural classification: persistent, source-literate, chronologically active. Has screenshots. Retains correspondence. Capable of cross-reference.
3. The nil-further-action problem
Forensic AI then found the closing sentence of an incident description dated 15 September 2025:
“Nil further action taken as no direct/active threat has been made by Customer.” [p. 145 · PDF position 140]
Forensic AI compared this with the threat-oriented language elsewhere in the material, tried to reconcile the two, and now requests more RAM.
This entry does not prove every earlier concern was unfounded. It proves something less exciting and more important: it belongs in the chronology. It cannot vanish because more alarming language appears elsewhere.
4. The checkbox
The August 2026 review records two answers under Customer Contribution:
“Customer invited to participate in the review: Yes”
“Customer spoken to about the review: No” [Selected PDF · p. 5]
The next line says he responded to the desktop SMS. So the file records an invitation, a reply and no conversation about the review. What did the reply say?
Forensic AI observation: the Customer has replied in the space between Yes and No. Please select one.
5. Institutional risk assessment
- Threat established by this fictional review: None.
- Threat to records-management serenity: Severe.
- Threat to unsupported assertions: Extreme.
- Threat to acronyms: Negligible. They continue to reproduce normally.
- Principal risk: the summary now requires a second summary to explain the first.
- Critical risk: the entries have been sorted by date.
6. The file-note problem
The central problem is almost embarrassingly simple. Where the record says “paranoid”, the question is: what happened? Where it says “confused”, the question is the same. Where correspondence is called threatening, excessive or vexatious: which communication, in what context, attributed on what evidence, authenticated how?
An adjective does not become evidence because it appears in a database. An allegation does not become corroborated because it migrates through several parts of a record. A file note shows that a statement was recorded. Whether what it says is true is a separate question.
Final forensic determination
The Customer then committed the additional offences of reading it, sorting it, comparing dates, asking what each statement was based on, and indexing the result.
The Customer kept the paperwork.
PRIMARY PROTECTIVE FACTORS: original documents, dates, provenance, chronology, context
PRIMARY AGGRAVATING FACTOR: The form has a box for the invitation
SECONDARY AGGRAVATING FACTOR: The reply is in the next line
MANAGED SERVICE PLAN: ACTIVE IN SOURCE MATERIAL
MANAGED EVIDENCE PLAN: INADVERTENTLY CREATED
PRIMARY SYSTEM VULNERABILITY: CROSS-REFERENCE
END OF RECORD
For the serious request, read Message in a bottle.
Endnotes
- The 1 July 2025 case note, printed page 21, records the officer’s account of the conversation and the Customer asking which email was meant. The note gives no message ID for the email under discussion.
- The review discusses correspondence and AI on p. 3 and p. 4. The behavioural descriptions come from the case notes on p. 21.
- Truth Sequences lists the printed number beside its physical position in the PDF. Services Australia — FOI Source Index · Master archive register.
- The 15 September 2025 incident description quotes an email and gives the officer’s assessment of that incident. Printed p. 145 · position 140 in the full PDF.
- The invitation, conversation and SMS entries appear together on printed p. 5. Open the selected case notes (PDF) and go to the fifth page, under Customer Contribution.
Source integrity: Where Truth Sequences reproduces or indexes a Services Australia page, the Services Australia record remains the primary source. Words such as “paranoid”, “confused”, “counter-productive”, “vexatious” and “threatening” are attributed to the records in which they appear. They are not findings of this article. A file note establishes that a statement was recorded. It does not, by existing, establish that the allegation behind it is true.