How They Falsify the Public Record

In a functioning democracy, the record of a public body's deliberations is the record. Board minutes are the mechanism by which the public, the parliament, and the auditors know what happened. When the record is accurate, it enables accountability. When it is not, it prevents it.

At Digital Health and Care Wales, the published record of board meetings contains approximately 10 to 15 percent of what was actually said. The remaining 85 to 90 percent is deleted before publication.

This is not compression. It is curation. And when CareNHS reconstructed the full record — from 2.6 million words of verbatim transcript across 37 board meetings — the pattern was unmistakable: what survives publication is disproportionately positive. What is deleted includes the warnings, the admissions, the financial figures, and the dissent that would have told the public the organisation was failing.

Published minutes showed a compliant board. Verbatim transcripts showed the challenger. The question is which record the auditors read.

What Gets Deleted

Across 37 board meetings from April 2021 to March 2026, CareNHS has documented 107 specific instances of substantive content removed from the published record — across 26 of those 37 meetings. Seventy percent of all board meetings contain sanitised content. These are not editorial trims for brevity. They are the removal of material that, had it survived, would have changed the public's understanding of what the board knew and when.

The very first DHCW board meeting — 1 April 2021, the founding meeting — has no published minutes at all. The meeting where the CEO said there were "no outstanding issues," where three executives held Professor of Practice titles undeclared, where inherited technical debt was concealed — exists only in the video record that CareNHS transcribed.

Warnings from independent members were disproportionately deleted. Rowan Gardner — the most technically competent independent member, who served from founding to March 2026 — issued 21 documented warnings. Her challenge that the 70/30 split between maintaining legacy systems and building new ones was "antithetical" to effective delivery was stripped from the published minutes. Her warning that "we will not get out of this position by continuing to do the same thing" was deleted. Her observation that technical debt is "as real as waiting lists" was left on the record — but the board took no action.

Disclosures to external regulators were erased. At his final meeting as Chair in September 2025, Simon Jones disclosed that he had written to the Auditor General of Wales about concerns regarding the Welsh Government's role in DHCW's governance. This disclosure — the outgoing Chair of a Level 3 organisation escalating concerns to the head of the national auditor — was deleted from the published minutes. The public record contains no trace that it happened.

Financial data was stripped wholesale. The September 2025 board meeting — held while DHCW was at Level 3 — had more than 20 specific financial figures deleted from its published record. Not rounded, not summarised. Removed entirely. At an earlier meeting, the NHS Wales system-wide deficit of £633 million was mentioned in board discussion — the figure was stripped from published minutes. In July 2025, staff burnout had increased 3.9% year-on-year; the specific figure was replaced with "Particular challenges around burnout were highlighted."

The Chair's own warning on structural risk was deleted. Jones warned that making "recurrent savings from non-recurrent" sources would "heap misery on misery every year." This described precisely the mechanism — vacancy savings as a structural dependency — that drove the organisation toward escalation. The warning was deleted. Thirty months later, independent member David Selway explicitly connected the vacancy savings mechanism to the escalation on the record — the first board member to state the causal link. That was deleted too.

The Director of Strategy admitted he could not fix the problems. In March 2025, Evans said: "We haven't fixed all of those issues in the last 12 months. And I'm not sure we can fix them without the support of Welsh Government." The man Welsh Government had seconded to implement their digital strategy, admitting he cannot fix the problems without the government that caused them. Deleted.

The Content That Survives

The deletions are not random. The content that survives publication is systematically skewed toward the positive, the reassuring, and the congratulatory. The content that is deleted is the content that would trouble a reader.

Ruth Glazzard articulated a transparency principle at a 2021 board meeting: "Every time we have a private discussion, the first question is: does this really have to be private?" This statement was deleted from the published minutes. What survived that meeting was the celebration.

Ifan Evans — the Director of Strategy, seconded from the Welsh Government — made 65 separate admissions across three years, each slightly more candid than the last. In July 2024, he told the board: "We have historically tended to be optimistic, and that is because we want to please very many of our stakeholders." It is the most revealing single sentence in 2.6 million words — the organisation's defining cultural flaw admitted openly by the person responsible for strategy. It was deleted.

"We have historically tended to be optimistic, and that is because we want to please." The most revealing sentence in 2.6 million words. Deleted from the published record.

The CFO, Claire Osmundsen-Little, tracked a VAT liability from September 2022 — when she was "reasonably confident" of recovering £5 million — through to January 2024, when she said it was "going to be on my tombstone." That quote was deleted. By February 2026, the unresolved exposure was £2.8 million.

In September 2024, Sam Hall warned that "time is running out on the existing platform" and staff on fixed-term contracts were "very concerned." The minutes retained the programme update. They removed the urgency and the human dimension.

They Prefer to Talk in Private

The sanitisation of published records is one mechanism. The preference for private discussion is another. Across the 37 board meetings, a pattern emerges: the more consequential the topic, the more likely it is to be handled behind closed doors.

Self-censoring live on camera. In July 2023, Evans prefaced a funding warning with: "I know this is a public meeting, but I think next year, year two and year three, there is less assurance around the DPIF funding." He flagged to the room that he was about to say something he would rather not say on camera. The candour he partially suppressed was then deleted from the published minutes anyway. In September 2022, he disclosed Welsh Government accountability conditions on the IMTP but stated they were "not for discussion today." Conditions placed on DHCW by its funder — raised at a public meeting, then immediately deferred.

"I know this is a public meeting, but..." He warned the room he was about to be honest. The honesty was deleted anyway.

Private by design. After Level 3 escalation, Jones proposed a private reflective session: "Could we have avoided getting here? It won't be a public meeting for obvious reasons." The Chair's instinct, when the organisation reached crisis, was to discuss it behind closed doors.

The Laboratory Information Management System — later declared infeasible after 53 months — was approved in DHCW's first-ever private session in September 2021. No financial details were shared publicly. The programme that would consume five years of board attention and end in failure began behind closed doors.

In November 2022, an independent member disclosed that the private session had produced a "very strong discussion about the consequences around obsolete and legacy systems." The substance stayed private. The public heard only that there had been a discussion.

The hidden risk register. At peak, 41% of corporate risks were hidden from public view — nearly half the risk register existed only in private sessions. In May 2022, 9 of 22 risks were private. Jones, in January 2025: "We held a private board meeting earlier today... when we are considering things like cybersecurity, it makes all sorts of sense to do that in private." By July 2025, five risks were reviewed only in private sessions.

Silence-as-consent. At Meeting 3, Thomas presented four new director roles. The board approved in silence. Jones: "I'm taking silence as consent." The roles created in that silence were the same roles whose holders would admit programme failures, describe data centre "never events," and face Level 3 escalation. The Kainos £20 million framework was approved in 15 seconds. The Compassionate Leadership Pledge was rubber-stamped in 15 seconds — at the same meeting where 65% burnout was buried.

The Welsh Government was complicit. In November 2024, Evans reported that Welsh Government had challenged why diagnostics programmes were rated red: "Government was asking... why then are those programs rated red amber?" WG pressured DHCW to soften its RAG ratings. By March 2025, the same Welsh Government escalated DHCW to Level 3 partly for programme delivery failure. The funder pressured the organisation to report more optimistically, then punished it for the consequences.

The Language Machine

Where content was not deleted outright, it was softened.

The CEO, Helen Thomas, told a board meeting that a data centre failure "should be a never event." In the published record, this became: "expressed appreciation for the transparency." The warning was inverted into a compliment.

When independent member Alistair Neill expressed "initial disappointment and frustration" at the Level 3 escalation — a measured criticism from a board member who had not been told escalation was coming — the published record replaced it with "Positive Engagement."

Evans's admission that a six-year strategy presentation was "a dangerous thing to do" was inverted to positive framing.

Thomas's statement that the LIMS programme was "causing some anxiety" was softened to "issues."

The pattern is consistent: specific, candid, attributable language is replaced with vague, positive, unattributable summaries.

The reader of the published record encounters an organisation that is reflective, transparent, and making progress. The reader of the verbatim transcript encounters an organisation whose directors know it is failing and say so — into a record that is then curated to say the opposite.

The AI Upgrade

From May 2025, the curation acquired a new tool.

Microsoft Copilot now generates meeting summaries that systematically soften language. But the AI output is not published automatically. A human editor reviews, edits, and approves the Copilot draft before publication. The AI produces the first draft of the sanitised record. A person approves it.

This is a two-stage curation process: algorithmic compression that strips specificity and softens tone, followed by human editorial selection that determines what survives. The result is a published record that is further from the verbatim transcript than at any point in DHCW's history — not because the AI is worse at recording, but because it is better at smoothing.

The February 2026 Programme Delivery Committee illustrates the endpoint. The Copilot-generated minutes retained 11.2% of the meeting — 3,500 words from 31,150. Every independent member challenge was erased: Gardner's escalation challenge, Hurle's clinical testimony, Evans's enforcement admissions, Hall's structural explanation. Only executive presentations and RAG ratings survived.

At that same meeting, Medical Director Hurle gave the most vivid clinical testimony in the entire record — describing patient record fragmentation in terms any patient would understand: "Waitrose has 421 stores across the UK. They have the same till in every shop... To see a patient, I now needed seven sets of notes because wherever that patient had been." All erased from the Copilot minutes. The most human, most memorable description of the system failure was the one the AI removed.

The Copilot minutes for May 2025 carry the footer: "DHCW SHA Board Meeting 28 November 2024 — Minutes produced with the support of Co-Pilot." The meeting was in May 2025, not November 2024. The AI that produces the sanitised record cannot even get the date right.

The organisation has automated its own information control.

Beyond the Boardroom

The curation of the public record does not stop at board minutes. It extends to every public communication the organisation produces.

The CEO's quarterly update

In January 2026, ten months after DHCW was placed at Level 3 Enhanced Monitoring, Helen Thomas published her public quarterly update to stakeholders. It opened:

"As we welcome in the new year, the pace of digital transformation across NHS Wales continues. We have ambitious plans for digital and data services over the coming 12 months which will bring benefits to patients, clinicians and the wider health system."

The update does not mention Level 3, Enhanced Monitoring, or the nine programme failures named in the Welsh Government's own Escalation Framework. It does not reference the Remit Letter, the 42% milestone delivery rate, or the £49 million in external contracts. It presents progress metrics without the remedial context that produced them.

A stakeholder reading this update — a GP, a hospital trust, a member of the public — would have no way of knowing the organisation was under government intervention.

The Annual Quality Report

The Annual Quality Report 2025 — the formal compliance document for the year Level 3 was imposed — contains zero references to Level 3, Enhanced Monitoring, or any of the nine named programme failures.

The governance statement

On page 1997 of DHCW's 2024-25 Annual Report, the Governance Statement reaches the sentence recording why the Welsh Government placed DHCW under Enhanced Monitoring:

"The rationale for the decision to increase the escalation status was the ongoing challenges with pace and delivery on key national priorities, including:"

Nothing follows. The next rendered line begins a new subject. In a 206-page governance document, the colon is unresolved. The sentence that should name the specific failures is a blank.

A 206-page governance document. One blank sentence. The rationale for government intervention ends at a colon with nothing after it.

The Data They Never Collected

There is a category of information control more effective than curation: never collecting the data in the first place.

Between 2018 and 2026, DHCW published no data that would allow anyone — the Senedd, the Welsh Government, the public, or its own staff — to assess whether the culture described as "the antithesis of open" by the Public Accounts Committee in 2018 had changed.

Zero whistleblowing data. No annual figure for disclosures received, their categories, or their outcomes. No record of whether those who raised concerns experienced detriment.

Zero disciplinary data. No information on proceedings initiated, categories, outcomes, or the seniority of individuals involved.

Zero leavers analysis. No data on why staff left, whether departures concentrated in particular teams or grades, or whether exit interviews indicated cultural problems.

Zero programme-level expenditure. Six of nine programmes under Level 3 escalation have no published expenditure figure.

In England, the National Guardian's Office has collected and published speaking-up data for over nine years. In 2024-25, Freedom to Speak Up Guardians across the English NHS handled more than 38,000 cases. Wales has no equivalent reporting mechanism. The absence is not an oversight. It is a policy choice. The data does not exist because the organisation chose not to collect it.

The most effective form of information control is ensuring there is nothing to suppress.

The buried review

In 2024, DHCW commissioned Atos Healthcare Consulting to conduct an independent stakeholder review. The organisation paid £207,100 for the work. The resulting report was based on 292 survey responses and 30 semi-structured interviews.

Only 13.3% spoke highly of the organisation. 50.4% said DHCW did not understand their work. 43.2% found it challenging to work with DHCW. The most common descriptors were "Slow," "Inconsistent," and "Confused."

DHCW did not publish the report. The 292 people who gave their time were not told the findings. The report was marked "for external use" and kept internal. It was obtained only through a private citizen's Freedom of Information request.

£207,100 to hear the truth. Then they buried it.

The Website Block

In 2026, CareNHS discovered that carenhs.org — this website — is blocked on the NHS Wales network managed by DHCW. Not flagged with a warning. Not slowed by a filter. Blocked entirely — classified alongside malware, phishing operations, and pornography.

Carenhs.org publishes articles built on publicly available sources: Welsh Government statements, Senedd committee transcripts, Audit Wales assessments, and DHCW's own published admissions. Every claim is sourced. Every organisation named is offered a right of reply.

The organisation under scrutiny has blocked its own staff from reading that scrutiny. Staff working inside an organisation under Level 4 government intervention cannot access the most comprehensive independent analysis of why it is under intervention.

The block sends three messages. To staff: do not look at criticism. To the public: the organisation that publishes Compassionate Leadership pledges treats accountability journalism as a threat. To the Welsh Government: the body it funds to serve the public is using public infrastructure to prevent the public from reading about it.

CareNHS has submitted a formal Freedom of Information request seeking the date of the block, who authorised it, what classification was applied, and what assessment — if any — compared the site's content against the criteria for that classification.

How We Found Out

The record was curated. The data was not collected. The review was buried. The website was blocked. So how does any of this become visible?

The verbatim transcripts

DHCW's board meetings are live-streamed. CareNHS downloaded, transcribed, and analysed every meeting — 37 board meetings and 24 committee meetings, approximately 2.6 million words of verbatim transcript. We built a knowledge graph of 1,067 entities and 2,522 evidenced relationships, each traced to a specific passage in a specific transcript.

The graph did what no human reader could: it read all 37 meetings simultaneously. It tracked the same warning raised by the same person at the same meeting cycle, year after year, deleted from the record year after year. It traced a single financial mechanism — vacancy savings — from a £300,000 surplus at month one to a structural dependency that the CFO designed into the financial plan. It identified the point at which the Chair stopped warning and started managing — and mapped the governance shift to the meeting where she took the chair.

What the graph reveals is not a single failure. It is a trajectory that was visible from month one, charted at every meeting, and erased from the published record at every publication.

The FOI trail

When DHCW's published accounts showed £757,000 in consultancy spending, CareNHS filed a Freedom of Information request. The response disclosed 49 contracts totalling £8.94 million — twelve times the published figure. We flagged the gaps. The next FOI revealed the total had grown to £49.1 million. DHCW's own caveat confirmed the number was still incomplete.

£757,000. £8.94 million. £49.1 million. Three questions. Three answers. Each technically accurate. Each concealing the next.

Three FOI requests. Three figures. Each one technically accurate. Each one concealing a number many times larger behind a classification boundary that a citizen would need to know existed before they could ask about it.

The registers produced by the three FOIs do not agree with each other. Contracts that appeared in the first disclosure are absent from the third. Either the classification changed or the registers are unreliable. Both possibilities undermine any single disclosure as an authoritative account.

What the Pattern Means

Nine governance instruments. In each case, the form is present and the substance is not.

Every disclosure meets the minimum legal standard. None produces the transparency the standard was designed to create.

This pattern is not explained by incompetence. The omissions are too consistent, too structurally advantageous, and too precisely aligned with the organisation's institutional interest to be accidental. A blank where the rationale should be. A report that costs £207,100 and is never published. A website blocked the moment it becomes inconvenient. These are the actions of an institution that has learned how to satisfy the form of transparency while defeating its purpose.

Every disclosure meets the minimum standard. None produces the transparency the standard was designed to create.

CareNHS welcomes a response from DHCW, Audit Wales, and the Welsh Government to the matters raised in this article. No response has been received to date. If received, we will publish it in full. Contact: carenhs@carenhs.org

This article is based on: systematic analysis of 37 DHCW public board meeting transcripts (April 2021 to March 2026), approximately 2.6 million words; a knowledge graph of 1,067 entities and 2,522 evidenced relationships; DHCW's published Annual Reports 2021-22 to 2024-25; the CEO's quarterly stakeholder updates; DHCW FOI responses FOI/5854154, FOI/6695599, and FOI/6695741; the Atos Stakeholder Review 2024 (obtained via FOI/6195807); DHCW's declarations of interest register and audited Related Party Transaction disclosures; and direct testing of carenhs.org accessibility from NHS Wales network-managed devices.

Rose Davies, CareNHS

Related: The Number That Keeps Growing — three FOIs, three numbers, each concealing the next