A Minister Was Questioned in Parliament From His Published Diary — 54 Meetings on the Record. Nine Months Later, NSW Republished It With 57, and No Note.

On 28 August 2025, the chair of the NSW Legislative Council’s Portfolio Committee No. 6, Cate Faehrmann, opened a line of questioning at budget estimates with a document in front of her: “I just wanted to turn to your ministerial diaries. On 31 July last year you had a meeting with Listen Out Sydney. Who was at that meeting?”

The minister at the table was John Graham — Special Minister of State, Minister for Transport, Minister for the Arts, and Minister for Music and the Night-time Economy. The document was his published diary disclosure for July–September 2024, a 54-meeting summary the government had put online the previous October. Faehrmann was pursuing who had lobbied for the Contemporary Music Festival Viability Fund — an emergency program announced on 17 September 2024 with an initial $3 million, whose first round ultimately paid $2.25 million to five festivals at up to $500,000 each, the recipients not made public until April 2025: “Who did you meet with before that viability fund was announced to discuss the fund and agree to the fund?” And, later in the same hearing: “I am looking at the integrity of this fund, Minister.”

Graham pointed her back at the document. “You’ve got my diary there, Ms Faehrmann, and all these things have been disclosed.” Asked directly — “Do you list all of those meetings in your diary?” — he answered: “I comply with the diary disclosures very directly and meet those requirements. Obviously I may well run into people who put views to me as I’m attending music events, but my engagement with the sector is quite extensive on this.”

On 4 June 2026, 280 days after that exchange, the government republished the diary she had been holding. The new version has 57 meetings. One of the three additions is dated 4 September 2024: a meeting with the Australian Recording Industry Association, recorded as “Portfolio matters – Music & Night-time Economy” — the same portfolio, inside the pre-announcement window the chair had been asking about. Nothing on the disclosure pages notes that the record changed.

This analysis makes no claim that the minister knew the published record was incomplete when he answered, and no claim about why any entry was added — the record does not show who changed the file or for what reason, and the memorandum governing the scheme has carve-outs that leave genuine judgment calls about what must appear. What the record does show, reproducibly, is narrower: the diary a parliamentary committee used to test a minister’s meetings was changed 19 months after publication, the change added a meeting inside the very window the committee had probed, and no version of the record says so.

The clock is almost never missed#

The disclosure scheme itself runs with unusual precision. Premier’s Memorandum M2015-05 — mandatory, in force since 2015, updated as recently as February 2025 — requires every minister to publish quarterly summaries of scheduled meetings “with stakeholders, external organisations, third-party lobbyists and individuals”, naming the organisation and the purpose — and, where a third-party lobbyist attends, the lobbyist, “the names of all individuals engaged by the third-party lobbyist to undertake the lobbying who are attending the meeting”, and the client. The timing rule is a single clause: summaries are to be published “one month after the end of each quarter.”

That is what happens. Of the 257 diary PDFs published on the disclosure pages since January 2024 — covering quarters from mid-2023 through June 2026 — 245, or 95.3%, carry server timestamps 30 to 32 days after their quarter ended. Whole quarters land in a single batch: all 23 files for April–June 2026 share the timestamp 31 July 2026. Measured on first publication, the scheme is close to clockwork.

Every ministerial diary PDF on the NSW disclosure pages uploaded since January 2024 (257 files, quarters 2023 Q3 – 2026 Q2), by days from quarter end to the file's server timestamp. 245 sit in the 30–32-day monthly batch. Of the 12 above the band, five are not late publications at all: they are records that were published in their batch and later changed.

The 12 files above that band are where the story is, because five of them are not late publications. They are revisions — records that went up in their quarter’s batch and were later republished or replaced with different content. Punctuality, the one thing the memorandum’s clock measures, is not the property that failed. What failed is a property the memorandum never mentions: that a published record, once relied on, stays what it was — or says when it stops being so.

Three diaries came back changed#

On the morning of 4 June 2026, within a five-minute window, three files were uploaded to the disclosure site, each covering an old quarter of John Graham’s diary and each carrying an “-updated” suffix in its filename — the only marker, on any surface, that these are second versions:

• July–September 2024, republished 612 days after the quarter ended and 581 days after the original was published;
• July–September 2025, republished 247 days after the quarter ended;
• January–March 2026, republished 65 days after the quarter ended.

The year pages now link the new files. The visible link label still names only the portfolio — its one on-page trace of the swap is the automatically generated file-size suffix, which went from 213.37KB to 371.17KB. The link’s title attribute reads “Minister Graham - Diary Disclosures - 2024 Q3” — the same text it carried before, minus a “.pdf” extension. And no page carries the words “updated”, “revised”, “superseded” or “corrected” anywhere in its visible text, its link labels or its link title attributes. Nothing on the page says the document behind the link is a different document.

What happened to the superseded versions is stranger still. The 2025 Q3 and 2026 Q1 originals were deleted — both URLs now return 404, and the 2026 Q1 original demonstrably existed as late as 11 May 2026, when the Internet Archive captured it. The 2024 Q3 original was not deleted: it is still online at its 2024 URL, returning HTTP 200, byte-identical to the earliest archived copy of it — captured 23 November 2024 — and still carrying its original 31 October 2024 timestamp. Reachable, citable, and marked by nothing as superseded. The same upload batch produced both outcomes: two old versions erased, one left standing indistinguishable from a current record.

The original 54-meeting version was the linked public record for 581 days — including 28 August 2025, when a budget estimates chair examined the minister from it. The 57-meeting republication of 4 June 2026 replaced it in the page's links; the original file remains online at its old URL with nothing marking it superseded.

Because the 2024 Q3 original survives, that pair can be compared row by row. A deterministic parse of both PDFs — committed alongside this piece, with verification gates that refuse to run if the reproduction fails — finds 54 dated meeting rows in the original and 57 in the republication. Three meetings are new: a 21 August 2024 meeting with the Australian Local Government Association (published with an empty purpose field), a 22 August 2024 lunch and speaking event listing 15 organisations from CPB Contractors and John Holland to Walker Corporation and Western Sydney International Airport, and the 4 September 2024 ARIA meeting. Five counterparty cells changed, three of them recording a different entity: a meeting recorded against the Association of Regulatory and Clinical Scientists — a clinical-sciences body — became a meeting with the Australasian College of Road Safety, a meeting recorded against a named individual became a meeting with Distillery Capital Pty Ltd, and Business Sydney became Business Events Sydney. And 25 of the 52 matched rows had their purpose fields rewritten — mostly a bare portfolio tag expanded into a specific subject (“Portfolio matters – Roads” becoming “Portfolio matters – Roads Act 1993 (NSW) Anti-Protest Laws” for a Unions NSW meeting), though in several rows the stated subject was replaced rather than expanded: a 19 August 2024 meeting with Australian Super moved from “Housing, Portfolio matters - Night-time Economy” to “Portfolio matters – Roads & Tolling”, and two gatherings recorded as portfolio meetings were reclassified as “Lunch & Speaking Event”.

Differences between the 54-row original (31 Oct 2024) and the 57-row republication (4 Jun 2026), from a deterministic row-level comparison of both PDFs. Categories overlap: a row with a changed counterparty can also have a rewritten purpose. 26 of the 52 matched rows are identical in both versions.

Most of these changes make the record more informative, not less. That is exactly why the absence of any notice matters: if the republished version is right, then the version Parliament and the public used for 19 months differed from it in 31 places — 28 rows changed and three meetings missing entirely — and nothing anywhere says which version governs, or that there are two.

The other two republished quarters cannot both be checked. The 2026 Q1 pair — the original recovered from the Internet Archive — differs in exactly one meeting-row field: a 12 March 2026 group meeting gained the name of an individual attending for a third-party lobbying firm. M2015-05 itself requires “the names of all individuals engaged by the third-party lobbyist to undertake the lobbying who are attending the meeting”, so the change reads as a required field being filled 65 days after the quarter closed. The 2025 Q3 original, though, is gone from both the live site and the Internet Archive, which has no capture of its URL. A republication of that diary exists; what it changed, if anything, cannot be established. Where this piece says “three diaries were republished”, it can only say of two of them what the republication did.

The estimates hearing used the old version#

The reason a changed old record is more than housekeeping is that these documents get used — and this one was used, on the record, in a formal accountability process, while it was incomplete.

The transcript of the 28 August 2025 hearing shows the committee chair working directly from the published diary: quoting the 31 July 2024 Listen Out meeting, the 3 September 2024 Australian Festival Association meeting, and pressing on who had access before the festival fund was announced. It shows the minister referring her back to the published record as the answer — “You’ve got my diary there” — and answering, when asked whether all such meetings are listed, that he complies with the requirements, while allowing that he “may well run into people who put views to me as I’m attending music events”.

The record she had could not show her a 4 September 2024 meeting with the recorded purpose “Portfolio matters – Music & Night-time Economy”, because that entry was not published until 4 June 2026. Whether that meeting would have drawn a question is unknowable. What is documented is that when this record was put to its stated purpose on the parliamentary record — public scrutiny of who meets a minister and why — the reader was working from a version the government itself later treated as needing correction, and was never told.

NSW parliamentarians use these diaries this way routinely. The parliamentary record over the window of this analysis shows a member telling the chamber she “reviewed Minister Kamper’s ministerial diary disclosures prior to budget estimates” (7 May 2025); a member using the Corrections Minister’s diary as evidence that a meeting did not occur — “there is not one listed on his public diary disclosures” (20 November 2025); and a 2024 motion in the Legislative Council about the Premier’s diary entries, in which another member quoted ICAC’s 2021 finding that “the published summaries of ministerial diary disclosures are not sufficiently detailed or meaningful for the public to understand who is meeting whom and why” (23 October 2024). The diaries are not an archive. They are a working instrument of scrutiny — which is what a version that can change without notice undermines.

Two other events sit close to the republication in time, stated here for what they are and no more. On 27 May 2026 the Legislative Council censured Graham over the findings of a committee inquiry into a grants program, and on 28 May he was suspended for the remainder of that sitting day over the government’s non-compliance with an order for papers concerning a review of hate-speech laws. The diaries went up on 4 June — eight days after the censure, seven after the suspension. Nothing in the Minutes of 27 May 2026, the Hansard of 28 May 2026, or the committee report the censure relied on mentions ministerial diaries at all, and the orders for papers made that week concern other subjects, with no diary component. The dates are reported here as dates.

Two more files changed without even a new filename#

The “-updated” suffix at least leaves a fingerprint. Two other diary files changed with no new filename at all — the same URL serving different content on different dates, detectable only because each file’s server timestamp had drifted from its upload batch and the Internet Archive happened to hold an earlier capture.

The Treasurer’s July–September 2025 diary was published in its quarter’s batch on 31 October 2025. Its live file now carries the timestamp 6 February 2026. Comparing the live file against the Internet Archive’s capture of 14 November 2025: all 74 meeting rows are present in both, and exactly one field differs. In the disclosed client list of a third-party lobbying firm attending a 22 September 2025 meeting, the name “Droneshield Ltd” — an ASX-listed defence technology company — appears in the November capture and does not appear in the live file. Nothing else changed.

Minister Harris’s October–December 2023 diary was published in its quarter’s batch on 31 January 2024; its live file is stamped 9 August 2024. Against the Archive’s capture of 19 May 2024, the live version adds an attendee (“Member for Tamworth”) to a 30 November 2023 meeting row — and introduces two typos: a date rendered “7 Noc 2023” where the capture reads “7 Nov 2023”, and “Hunter and Central Coast Development Corporation” rendered “Huner and Central Coast Development Corporation”. The change process, whatever it was, degraded the record it touched.

Both changes were verified at the level of decoded bytes and extracted text, with a control: the one Graham file whose timestamp matches its batch is byte-identical to its own Internet Archive capture, so the comparison pipeline reports no difference where none exists. Both files’ original batch-date publication is verified the same way — the Archive’s captures record the origin server still reporting 31 October 2025 and 31 January 2024 timestamps before the replacements. But the method’s reach is short. It can only see a replacement where the timestamp moved and a pre-dating capture exists. Most diary files have no capture at all. Two is therefore a floor, not a count — and this piece makes no claim about how many other files have changed invisibly, in either direction.

No claim is made, either, about why these two files changed, who changed them, or whether the changes corrected errors or introduced them. The Droneshield entry’s removal could be the deletion of a mistaken disclosure or the loss of an accurate one; the record offers no way to tell, and that is the point. A disclosure scheme whose records can change at a fixed URL, without notice, offers a reader no way to tell either.

No rule says any of this is wrong — which is the finding#

The governing memorandum’s text was read in full for this analysis. M2015-05 requires publication, sets the one-month cadence, lists what a summary must contain and what may be withheld. Its text contains no provision — none — about correcting an already-published summary, versioning one, noting a change, retaining a superseded file, or marking one superseded. The words “correct”, “amend”, “revise”, “version” and “error” do not appear in it. It names no compliance mechanism beyond its own mandatory status, and no auditor. A search for any NSW policy on correcting published diary disclosures — across the memorandum, the Cabinet Office disclosure pages, the Ministers’ Office Handbook chapter on records, and ICAC’s report on the scheme and the government’s response — found nothing on the question.

The nearest thing to a standard is the State Records Act 1998, which obliges ministerial offices, as public offices, to “make and keep full and accurate records” of their activities. Correcting a wrong disclosure is consistent with that duty — arguably required by it. The gap this analysis documents is not that records were corrected. It is that every correction found here was made with no notice, no changelog, no consistent treatment of the superseded version, and no way for a past reader — including a parliamentary committee — to learn that what they read has been overtaken.

ICAC saw most of this scheme’s weaknesses five years ago. Its June 2021 Operation Eclipse report found the published summaries “not sufficiently detailed or meaningful for the public to understand who is meeting whom and why”, criticised the scheme’s “content, format and timeliness”, and noted — in a bullet that reads differently after this analysis — that the files “do not seem to follow consistent naming conventions.” It recommended diaries be published monthly, “in a single, searchable document or database”, with each minister’s name shown, and that the requirement move from a Premier’s memorandum into regulation with “the lobbying regulator … responsible for compliance.” In October 2022 the government of the day marked the first of those recommendations — monthly publication, a single searchable database, each minister named — “Accepted”, though its stated response promised only to “introduce a requirement for all Members of Parliament to publish diary disclosures on a routine basis”, and it marked the regulation-and-compliance recommendation “Accepted in principle”. As at August 2026, the disclosures remain quarterly, per-portfolio PDF files, governed by a memorandum with no compliance mechanism — a format in which a record can be republished, replaced or deleted without anyone being told. ICAC’s diary recommendations address what the summaries say, their format and their timing — not whether what they said yesterday still stands. The report does prescribe a challenge-and-correct mechanism for the scheme’s sibling dataset, the Lobbyists Register, which makes the silence on diary corrections the more striking.

What could explain this instead#

The revisions are the system behaving well, not badly. On this reading, a minister’s office found errors and omissions in old disclosures and fixed them — late honesty being better than none — and the State Records Act’s accuracy duty points exactly that way. This explanation is fully consistent with the data, and the piece does not dispute it. It does not touch the finding, which is about the absence of any marker that a correction occurred. A corrected record and a changed record are indistinguishable to the reader unless someone says which it is; nothing here says.

The added meetings may never have been required disclosures at all. M2015-05 excludes meetings that are strictly personal, electorate or party political, social or public functions, and anything with “an overriding public interest against disclosure” — though it qualifies the functions exclusion: a non-public substantive discussion of a portfolio matter at such a function should still be disclosed. Two of the three added meetings — a 15-organisation lunch billed as a “Lunch & Speaking Event”, and a meeting published with an empty purpose field — could plausibly have been judged out of scope in 2024 and in scope in 2026. If so, the original was not a breach, and this piece does not claim it was. The republication would then reflect a change of judgment about the same facts — made, like everything else here, without notice.

The timestamps could mislead. The one-month wall and the revision dates rest on HTTP Last-Modified headers, which record a server file’s modification time, not a legal act of publication. A re-upload of identical bytes would move a timestamp without changing anything; a platform migration could rewrite timestamps in bulk. Three checks limit the damage: the timestamps cluster almost perfectly on month-end batches, which random drift would not produce; the three cases with independent Internet Archive corroboration all agree with their timestamps; and every content claim in this piece rests on a byte- or text-level comparison of two dated files, not on a timestamp alone. The residual risk runs in one direction — a changed file whose timestamp did not move would be invisible — and that widens, not narrows, the finding’s floor.

The two in-place replacements could be routine, or could be the lot. The detection method requires both a drifted timestamp and a pre-dating Archive capture, and most files have no capture. It is possible in-place correction is common and mostly benign; it is possible these two are the only ones. The data cannot distinguish, so the piece claims only “at least two”. Either way, the records changed and nothing marked them changed.

Coincidence of dates around the republication. The 4 June 2026 upload came eight days after a censure and seven after a suspension. A connection would be a story; none is established, the papers ordered from the minister concern unrelated subjects, and this piece asserts none. The adjacent explanation — that some unrelated internal review prompted a tidy-up of old diaries at that moment — is just as consistent with the record, and equally unverifiable.