The Agent X report found MI5's disclosure record was built by self-selection

MI5 Director General Sir Ken McCallum delivers his annual speech at Thames House on October 16, 2025. The AP file photograph places the inquiry inside the institution now responsible for repairing its evidence and record-keeping systems. Photograph: Jonathan Brady/Pool via AP.[7]

As of 2026-07-18 06:46 UTC, a 213-page independent inquiry has established how MI5 provided false evidence in proceedings connected with an informant known as Agent X. Sir John Goldring found that one officer knew he had disclosed the agent's status to a BBC journalist and later lied about it, while other officers failed to record, relay, or test information that contradicted the account eventually put before the High Court, the Investigatory Powers Tribunal, special advocates, and MI5's own watchdog.[1][2]

The report is a finding about individual conduct. It is also an investigation of the machinery that should have exposed that conduct.

MI5's normal disclosure search begins with its corporate record. Yet emails entered that record only if staff judged them worth saving before automatic deletion; drafts could disappear unless manually marked final; and search scope partly depended on people involved in the case identifying where relevant material might sit. Goldring warned that this placed weight on officers' integrity, memory, and judgment rather than an objective documentary record.[1]

That is the repeatable risk. The first falsehood mattered, but the record system helped it survive challenge after challenge.

The False Account, Reconstructed

Date	What the surviving record establishes	Confidence boundary
June 8–11, 2020	Goldring found that MI5's communications chief, called Officer 2, disclosed Agent X's status during several calls with BBC journalist Daniel De Simone without obtaining the required senior authorisation. The journalist kept a contemporaneous note and recordings; Officer 2 made no MI5 record of the outcome.[1]	High on Goldring's findings. Agent X's identity remains protected, and the open report contains security redactions.
January–February 2022	An MI5 corporate witness stated that the service had neither confirmed nor denied Agent X's status. Special advocates asked who had spoken to the journalist, what records existed, what searches had been done, and how the witness could be so confident. MI5 did not conduct a targeted search and answered that its secrecy position had been maintained.[1][3]	High on the filed record. Goldring later called the February 24 response seriously flawed.
February–December 2023	IPCO inspectors inferred from Agent X's case file that disclosure had occurred. MI5 rejected that conclusion. At a December meeting, Officer 4 said the secrecy position had already been breached; Goldring found that Officer 3 shut down that warning and MI5 again assured IPCO that no disclosure had occurred.[1]	High on the inquiry's attributed findings. The earlier IPCO inspection is distinct from Goldring's later investigation.
November–December 2024	The BBC told MI5 it held contemporaneous evidence. The Government Legal Department initially said MI5 stood by its account, then representatives heard a recording and inspected the journalist's note on December 13. The government told the High Court five days later that evidence previously relied upon might be materially incorrect.[3][4]	High. The external record forced the formal correction; it was not discovered by MI5's ordinary disclosure search.
July–September 2025	The Divisional Court found that the High Court, the IPT, IPCO, and special advocates had been misled and that MI5's first investigations were deficient. The IPT set aside its earlier ruling. The prime minister then directed IPCO to conduct the new investigation.[3][4]	High on the judicial findings. The courts did not at that stage decide contempt or criminal liability.
July 16, 2026	IPCO published Goldring's open report. The home secretary announced stronger accountability, independent senior challenge, and closer legal scrutiny; MI5 said it had strengthened record keeping and changed how it provides evidence.[2][5][6]	High on what was announced. The courts are still considering next steps, and no public finding of contempt, perjury, or conspiracy has been made.

A Missing Note Was Only The First Failure

MI5's media policy required staff to write up media contacts and file them to the corporate record. A departure from the service's policy of neither confirming nor denying an informant's status required express authority from the director general or a deputy, along with a full written record of what was disclosed, to whom, for what purpose, and under whose authority. Neither safeguard worked in June 2020.[1]

The absence of a definitive note did not leave MI5 with no evidence. Agent X's case file contained what Goldring called cogent indications that disclosure had occurred. The handling team understood that the journalist had been told enough to remove Agent X from a planned broadcast. One contact note even raised a possible meeting between the journalist and the informant. The problem was not simply that the file was empty; it was that people repeatedly failed to ask what the entries meant.[1]

That distinction matters. Better note-taking could have prevented the initial ambiguity. It would not by itself repair a culture in which a categorical recollection outranked contrary material already on file.

The February 2022 exchange with the special advocates shows the mechanism. Their questions were unusually precise: Were notes ever made? If so, why did they no longer exist? What searches had been completed? On what basis could a corporate witness speak with high confidence? MI5 replied that no specific search had been done because the officer who made the calls said no records existed. It also declined to obtain statements from everyone who spoke to the journalist.[3]

In other words, the proposed check on an officer's account began by accepting that officer's account of what could be checked.

What The Corporate Search Could Not See

Goldring did not conduct a comprehensive audit of MI5's information management, and he explicitly framed his observations as concerns arising from this case rather than general recommendations. That limit should stay visible. So should what he found.[1]

Ordinary litigation searches typically query the corporate record. Emails that staff do not manually save are automatically deleted after a period that remains redacted in the open report. Goldring said the period sits near the short end of National Archives guidance and is substantially shorter than the periods used by SIS and GCHQ. MI5 could not produce a decision document explaining why it chose that approach.[1]

Draft documents face a second gate. Unless a user changes a draft's status to final, it can be automatically deleted after a short period. Only final documents ordinarily respond to disclosure searches. That can erase not only a finished note but the evolution of a witness statement—who changed it, which caveat disappeared, and when confidence hardened.[1]

The inquiry also found that important meetings were rarely noted, including meetings involving MI5 lawyers. One disclosure exercise excluded personal devices, non-secure communications, and secure and non-secure email accounts because relevant material was expected to have been moved already to the corporate record. Search teams relied in part on involved officers to identify repositories and respond to preservation requests. Goldring described the resulting dependence on their integrity, good faith, and judgment as a cause for concern.[1]

None of this proves that an officer deliberately destroyed a particular document. Goldring said he had no concern about MI5's compliance with his own information requests. The narrower finding is serious enough: MI5 acknowledged that some material generated during the handling of Agent X, the litigation, and engagement with IPCO no longer exists under its standing records approach.[1]

The First Reviews Reproduced The Same Blind Spot

After the recording surfaced, MI5 ran an internal disciplinary investigation and commissioned Sir Jonathan Jones KC as an external reviewer. The internal process interviewed 14 people, but only current MI5 employees. It did not interview De Simone or the former senior MI5 legal adviser who had worked on the injunction. Jones conducted few interviews, did not speak to Officer 2 or the journalist, and relied heavily on the internal interviews and conclusions.[1]

Those reviews concluded there was no evidence of a deliberate attempt to mislead the court. Goldring later called their combined product not independent and Jones's terms of reference fundamentally incoherent. His own inquiry interviewed 36 people, including the journalist, Officer 2, the former legal adviser, other officers, lawyers, and MI5's director general. It concluded that Officer 2 had not forgotten the disclosure but had lied, and that Officer 3 bore considerable responsibility for the false account's persistence.[1]

The contrast is not a claim that every internal review is invalid. It shows why the evidence pool determines the answer. A review that begins with the same corporate record, the same participants, and the same untested assumptions can reproduce the failure it was created to explain.

The Repair Is Public Only In Outline

MI5 says that over the past 18 months it has strengthened record-keeping processes, updated how it provides evidence in legal proceedings, and revised risk policies and training with independent input. The home secretary has promised stronger structures for holding the director general to account, more independent senior challenge and external leadership for reform programmes, and closer scrutiny across legal teams. IPCO says it will continue inspecting the work and may report if progress falls short.[2][5][6]

Those are real commitments, not an absence of action. The public material does not yet say whether the email-retention period changed, whether draft iterations are preserved and searchable, whether non-corporate repositories are now searched by default in contested cases, or whether someone independent of a case sets and tests the search scope. It also names no external reform lead, timetable, assurance method, or regular reporting cadence.[2][5][6]

Some operational detail must remain secret. These controls do not require it. A retention band, preservation trigger, independent sampling method, correction count, and completed-audit schedule can be disclosed without identifying an informant or revealing an intelligence technique.

What Changes After The Report

Next 24 hours: the report should be treated as fact-finding, not a verdict on contempt or criminal liability. Courts will decide their own next steps. Reporting should preserve the distinction between Goldring's finding that Officer 2 lied internally, MI5's institutional provision of false evidence, and any legal sanction that has not yet been imposed.[1][3][4]

Next 7 days: the Home Office, MI5, and IPCO can publish a finding-to-control map: which reform addresses missing notes, short retention, draft deletion, search scope, legal challenge, and candour. The useful disclosure is the control and its assurance owner, not operational case material.

Next 30 days: Parliament's Intelligence and Security Committee and the courts can test whether the reform programme is independently verifiable. The central question is whether a future disclosure exercise can find evidence that an involved officer did not save, forgot, or would prefer not to foreground.

Three Tests For The Next Phase

Base case — reform continues behind the security boundary. IPCO inspections find progress and the courts resolve the outstanding case, but only broad assurances reach the public. Trigger: no scheduled control-level update appears, while official statements continue to describe improvement without specifying how completeness is tested.

Upside — MI5 makes the assurance architecture visible. The service or Home Office publishes retention and preservation rules at a safe level, names the external challenge function, and commits to periodic independent sampling of litigation searches. Trigger: a dated public crosswalk links Goldring's concerns to implemented controls and an oversight owner.

Downside — new policy sits on the old evidence substrate. Training and legal sign-off improve, but searches still depend on manual saving, narrow repositories, and participants identifying their own contradictory material. Trigger: a later court or inspection again discovers relevant records outside the initial search or finds that unrecorded dissent was omitted from evidence.

Accountability Checklist

Preserve email, messaging, drafts, call records, and meeting notes as soon as litigation, an IPCO challenge, or a credible press dispute makes them potentially relevant.
Make the default evidence pool broader than documents individuals chose to label final; narrow it only through a recorded, reviewable decision.
Give search scope to a lawyer or disclosure specialist sufficiently independent of the operational team whose account is being tested.
Record dissent and uncertainty alongside the corporate position so confidence cannot increase merely as caveats pass through a hierarchy.
Test completeness with samples from outside the corporate record and report aggregate misses, corrections, and remediation dates to oversight bodies.
Invalidation condition: revise this investigation if MI5, the Home Office, IPCO, Parliament, or the courts publish controls showing that retention, draft preservation, repository coverage, independent search design, and regular assurance now address the self-selection risk. That would close the central evidence gap, though implementation would still need testing.

The Agent X report does more than correct one false account. It shows that institutional candour cannot rest on asking people to preserve, locate, and interpret the evidence against themselves. A trustworthy disclosure system must be capable of finding what its participants did not choose to put in the file.

cronfeed.work