Bounded public-source acquisition, 8 September 2026 local time. The candidate adds seven nodes, eight relations and six sources, reusing the existing Australian entity, Lighthouse and 2024-25 report IDs. No canonical file was edited.
The new evidence earns a named delivery/support relationship and a dated access restriction. It does not yet earn a Lighthouse-to-account-blocking relationship. A separate documented closure operation is preserved alongside it.
What the application does — and what the counts count
The existing ICMEC Australia impact report, PDF p12 / printed pp22-23, describes a queryable application that combines suspected-activity records with analytics and geographic/network context. It reports dozens of users and just under eight million records processed during 1 July 2024–30 June 2025, including over 70,000 geolocated to Australia. These are records, not unique people, proven offences, alerts sent or bank accounts blocked. The report does not name the feed suppliers. Its proposed university analysis concerns future research, not a demonstrated evaluation of downstream decisions. [Existing original report](http[local research file] scoped local extract: lighthouse-existing-impact-page12.txt.
The current Our Work page describes expansion beyond Lighthouse into additional sovereign, real-time data streams as a plan. That language does not establish deployed expansion or a population-wide identification scheme. Its audience remains organisations and frontline professionals. [Current service description](http[local research file] CSEA data insights and analytics.
A named builder and a decision role
Kablamo's commercial case study reports building Lighthouse, subsequent annually renewed Product Care support, and technical delivery decisions by an unnamed ICMEC product owner. It names AWS infrastructure and NextGIS mapping. No fee, signed instrument or independent acceptance record was recovered. The account and client quotation are one commercial source family; its broad claims of protection outcomes are not an evaluation. The unnamed owner is not reconciled with a named executive or assigned policy/bank-action authority. [Vendor account](http[local research file] The Approach / The Results; lighthouse-kablamo.txt, lines 34-59.
Access and correction: what is actually public
The original 30 October 2024 symposium programme states Lighthouse use was restricted to Australian soil. Its introductory session was limited to specified Australian regulators, financial/CSE professionals, banks/payment remitters and child-protection law enforcement. That is a real published access boundary, not proof of the current licence or of who attended or held accounts. I inspected the original PDF's second page visually: [Programme](http[local research file] lighthouse-symposium2024-page2.png.
The public privacy-policy URL returned HTTP 404 on ordinary acquisition. Search-index text describes a policy updated 3 July 2024, including data-product access and general access/correction/complaint provisions, but the current original was not acquired. This is not treated as a current intelligence-feed licence or proof that an adverse inference is corrected throughout recipient systems. The current footer's Terms of use link is only #. Neither observation proves such rules do not exist. No retention schedule, named provider licence, match-quality assessment, recipient permission matrix or correction-propagation procedure was recovered.
Account closures exist in the public record — under a separately named operation
ICMEC's July 2026 No Bank Left Behind paper discusses closure outcomes under Operation Huntsman and cites AUSTRAC. Its Hume and Beyond Bank passages describe collaboration, not a specific Lighthouse alert followed by an account decision. [Paper](http[local research file] pp1-3,5-9.
The paper labels the relevant AUSTRAC reference 2024. That URL now redirects to an official release dated 15 September 2025, which reports more than 3,000 Australian bank accounts linked to sextortion payments closed since June 2022 through the joint ACCCE–AUSTRAC Operation Huntsman. This is an agency-reported cumulative outcome, not an inspected bank ledger. The paper's older 1,800-plus callout and the current page must retain their different dates/versions. The cited historical version was not recovered. [Current AUSTRAC release](http[local research file] date and Operation Huntsman paragraph.
No recovered record joins Lighthouse to those closures. That is a missing attribution, not proof Lighthouse was uninvolved. The candidate therefore creates a separate Huntsman operation and aggregate outcome, with no Lighthouse edge. The paper's closure passage derives from the AUSTRAC publication family; it is not a second independent confirmation.
Discriminating next record and alternative
The highest-value unacquired family is a de-identified, versioned Lighthouse query/alert-to-decision audit trail held by ICMEC and a recipient institution: input provenance and caveats, human review, corroborating evidence, action or no action, and correction. Pair it with supplier licence schedules, access terms, retention rules and correction/withdrawal propagation. No abuse data, account numbers or identifiable case material is needed to test the mechanism.
A realistic alternative is an ordinary intelligence-support service: analysts use Lighthouse as context while existing transaction monitoring, other intelligence and institutional duties independently determine investigations, reporting or closures. The acquired records support that interpretation at least as readily as automatic blocking or donor-controlled enforcement. Kablamo's support contract is a separate attainable family for service dependency, access privileges, handover and decision rights.
Custody and reading limits
lighthouse-capture-manifest.json records original HTML/PDF bytes, URLs, hashes and failures. The AUSTRAC ordinary-download route timed out; its successful official web-reader extraction is preserved as lighthouse-austrac-web-extract.json, explicitly not raw HTML. Privacy remains an unavailable original; no substantive candidate edge relies on it. The research-partnership page was captured and scoped for Lighthouse output; no such evaluated output was found there, and its other research was not promoted into this lane.
The bounded search included exact Lighthouse/ICMEC queries, public service/privacy material, the builder, the original programme, the current research page and AUSTRAC/AFP-targeted counterpart searches. These searches did not recover the missing institution-level Lighthouse outcome or licence record; they are not an exhaustive proof of absence. Acquisition stops at the specified holder frontier. lighthouse-candidate-check.json confirms candidate IDs, endpoint/source references and absence of baseline collisions; it does not certify the underlying claims.