What we found
In a consumer alert published 27 August 2026, the FTC says scammers know people want to donate after weather emergencies and natural disasters, and hope to steal money intended for disaster relief. The agency describes two versions of the same play: scammers create fake charities that seem official and claim to help people affected by disasters, or they impersonate well-known charity organizations so a familiar name draws the donation. The alert names recent flooding in Nepal and Tibet and recent earthquakes in Venezuela, Japan and Colombia among the events donors may be responding to, and warns that a donation request can arrive by text or be posted on social media by a friend. According to the FTC, a demand that you pay only by cash, gift card, wire transfer or cryptocurrency is a warning sign, because legitimate charities accept donations in many different ways. Before you give, stop and look the charity up yourself on Charity Watch or Give.org, and if the ask came by text, get the donation number from the charity's own website instead of the message. The alert does not report how many people have lost money to these fake charities, how much was taken, or which charity names have been impersonated. It also does not describe any particular group of donors as being targeted more than others, so we name none. LIVEFRAUD ALERTS is independent and not affiliated with the FTC or any charity or rating organisation named here; this alert was written only from the harvested source above.
- The single supporting row is an official FTC consumer alert, published with a date, describing the scam pattern in the agency's own words.
- Only one source was harvested, with no independent press or community reporting to corroborate it.
- The source is a general prevention advisory rather than an account of a specific incident, so no victim counts, losses or impersonated charity names are available.
- Reviewed by 3 models, 2 from independent houses.
THE RULE
Money sent to a fake or impersonated charity by gift card, wire or cryptocurrency is unlikely to be recoverable, and the ask often arrives at the moment you are most motivated to give.
What we don’t know
- How many people have been approached or have lost money to fake disaster charities.
- Which well-known charity names are being impersonated.
- Whether the text-message and social media approaches described come from the same operation or many unrelated ones.
- How the money moves once sent, and whether any of it has been recovered.
The bench — who voted
3 INDEPENDENT AI MODELS REVIEWED THIS. 2 OF 3 — ONE MODEL DISSENTED. READ WHY.
The card names a count. Here are the seats behind it, with what each one said.
ANTHROPICclaude-opus-5 · MODERATE
The single evidence row covers only two things: that scammers create fake charities or impersonate known charities after disasters, and that a cash/gift card/wire/crypto-only demand is a red flag. The finding adds a publication date (27 August 2026), a named list of triggering events (flooding in Nepal and Tibet, earthquakes in Venezuela, Japan and Colombia), delivery channels (text message, friend's social media post), and specific verification services (Charity Watch, Give.org, plus the advice to take the donation number from the charity's own site). None of that appears in the evidence supplied, so as presented these are assertions the reviewer cannot check.
DISSENT, PRESERVEDThe CLAIM itself — that the FTC warns of fake charities and impersonation of known charities intercepting disaster-relief donations — is fully carried by row [1] and should not be withdrawn. My objection is to the surrounding detail, not the core warning.
DISSENT, PRESERVEDThe alert's negative disclosures are exemplary and worth keeping on the record: it states plainly that no loss totals, victim counts or impersonated charity names are known, and it declines to name a targeted donor group rather than inventing one. That restraint should not be lost if the panel trims the piece.
DISSENT, PRESERVEDIf the extra detail comes from the same FTC alert and was simply not carried into the evidence row, the fix is to expand the evidence set, not to soften the finding. A reviewer working from one summary row cannot tell the difference between an unsourced embellishment and an under-recorded source, and that ambiguity is itself a process problem.
GOOGLEgemini-flash-latest · MODERATE
The finding includes specific details not found in the evidence row, including specific disaster locations (flooding in Nepal/Tibet, earthquakes in Venezuela, Japan, Colombia), communication channels (texts, social media posts by friends), specific research resources (Charity Watch, Give.org), and the publication date (27 August 2026).
GROQopenai/gpt-oss-120b · HIGH
Finding adds unsupported specifics: exact publication date, named disasters (flooding in Nepal/Tibet, earthquakes in Venezuela, Japan, Colombia), delivery methods (text, social‑media friend), and advice to use Charity Watch or Give.org, none of which appear in the cited evidence.
Reviewed by 3 independent models; all judged the finding to go beyond the evidence.
▼ Protocol & challenge record
Objection (medium): Verb stretch in the claim line: 'intercept money intended for disaster relief.' The source says scammers 'hope to steal money intended for disaster relief.' 'Intercept' implies diverting funds already in transit to a real charity — a different and more sophisticated mechanic than what the FTC describes (soliciting donations under a fake or borrowed name). Use the FTC's verb ('steal', or better 'divert donations to themselves').
Resolved: Replace 'intercept' with the source's own framing: 'to steal money intended for disaster relief' / 'so donations go to the scammer instead of people in need'.
Objection (medium): 'Two versions of the same play' is an inference the source does not support and it contradicts the draft's own unknowns list, which asks 'Whether the text-message and social media approaches described come from the same operation or many unrelated ones.' The FTC says scammers 'might' do one or 'might' do the other; nothing attributes both to one actor or one campaign. Either drop the unifying framing ('The agency describes two tactics') or drop the unknown. As written the finding asserts what the unknowns say is unknown.
Resolved: Rewrite to 'The agency describes two tactics' and keep the unknown about whether the approaches are one operation or many, which then no longer conflicts.
Objection (medium): Selective omission of the hazard list. The source's first-named disaster type is 'hurricanes'; the draft lists only Nepal/Tibet flooding and Venezuela/Japan/Colombia earthquakes. For a consumer audience this quietly relocates the risk offshore and makes the alert look less relevant to domestic readers, when the FTC's own examples include hurricane relief and 'wildfire recovery' as search terms. Add hurricanes/wildfires.
Resolved: Add hurricanes (and the FTC's 'wildfire recovery' example) to the list of disaster contexts named in the alert.
Objection (medium): The advice paragraph drops the FTC's own reporting and resource channels — ReportFraud.ftc.gov for charity imposters and ftc.gov/charity for guidance — and also drops the agency's first listed tip ('Do some research… search online for the cause') and its instruction to check how much of a donation actually reaches beneficiaries. For a fraud alert, omitting the official reporting route is a substantive loss of actionable content, not a stylistic trim.
Resolved: Restore the FTC's research-first tip, the 'how much reaches beneficiaries' check, and both official channels: ftc.gov/charity and ReportFraud.ftc.gov for reporting charity imposters.
Objection (low): Provenance labelling is wrong on the advice sentence: row_ids is empty, yet every element of that advice (Charity Watch, Give.org, confirm the number from the charity's website) is lifted almost verbatim from row 17186abd. Empty row_ids signals to a reader or downstream check that this is the publisher's own counsel rather than the FTC's. Attribute it to the row.
Resolved: Attach row_id 17186abd-680c-451f-940b-2494055cb802 to the advice sentence, since it is the FTC's advice verbatim, not the publisher's.
Objection (low): Internal contradiction between the limitation and the callout: the finding states 'It also does not describe any particular group of donors as being targeted more than others, so we name none,' while audience_callout names 'DISASTER RELIEF DONORS.' The source does address would-be disaster donors, so the callout is defensible — but then the 'we name none' sentence overstates the abstinence. Reconcile the two.
Resolved: Soften the limitation to 'the alert does not single out any subgroup of donors beyond people moved to give after a disaster', which is consistent with the DISASTER RELIEF DONORS callout.
Objection (low): Unsourced assertion in the risk_line: 'unlikely to be recoverable' and 'arrives at the moment you are most motivated to give' appear nowhere in the harvested row. The first is a widely held and probably correct generalisation about gift cards, wires and crypto, but the row does not say it; the second is pure editorial framing. Either mark as the publisher's own reasoning or cut.
Resolved: Either attribute the irrecoverability and urgency points as the publisher's own assessment, or drop them from the risk_line and keep it strictly to the payment channels the FTC names.
Objection (medium): Date currency: the alert is dated 2026-08-27 and its geographic examples are all flagged 'recent' by the FTC. The draft correctly attributes 'recent' to the alert, but does not state an as-of date for its own publication. If this ships materially later, 'recent flooding in Nepal and Tibet' reads as a live event report when it is a 27 August 2026 snapshot. Add an explicit as-of line.
Resolved: Add an as-of date for this alert and keep 'recent' explicitly inside the attribution ('the alert, dated 27 August 2026, calls these events recent').
Objection (low): QA failure in the generated option lists, even though nothing bad was selected. callout_options contains 'ATTENTION: SCAMMEDS' (victim-labelling and malformed) and 'ATTENTION: DISASTERS'; directive_options contains 'Send this to any disaster you know', 'Forward this to the disasters in your life' and 'Send this to any official you know'. targeting_dropped lists 'giving' and 'month' as candidate target groups. These are extraction artifacts, not audiences. The correct output here (share_directive null) was reached despite the generator, not because of it; the generator should be flagged.
Not resolved — preserved on the record.
Objection (low): Confidence is mis-scoped. The claim is a statement about what the FTC says, verified against the FTC's own consumer.ftc.gov page with a matching /2026/08/ URL path. Absence of independent press corroboration is not a real weakness for a claim of that shape — no third party is needed to establish that an agency published an advisory. The genuine limitation (no incident data, no named impersonated charities, no loss figures) is already captured separately. 'Moderate' understates attribution certainty while the confidence_reasons blur the two questions.
Not resolved — preserved on the record.
Objection (low): Entity precision: the source's 'Charity Watch' is CharityWatch, and Give.org is the BBB Wise Giving Alliance. Reproducing the FTC's spacing is defensible, but readers searching 'Charity Watch' can land on lookalike rating sites — a live risk in a piece whose whole subject is impersonation. Give canonical names and note Give.org's operator.
Resolved: Render as CharityWatch and Give.org (BBB Wise Giving Alliance), noting these are third-party raters the publisher is not affiliated with — the disclaimer already covers affiliation.
Objection (low): watch_icons 'bank' is a poor match for 'wiring money' and may read as bank-account compromise, which the source does not allege; cash and cryptocurrency — both named payment red flags — have no icon. Minor, but the icon set misrepresents the payment-channel mix.
Resolved: Swap or supplement icons so cash and crypto are represented and 'bank' is clearly a wire-transfer icon.
Preserved dissent
ON THE RECORDThe claim line's 'intercept' is not what the FTC said and should not survive to publication. Fake charities solicit; they do not intercept. This is the kind of small verb upgrade that turns an accurate advisory into a slightly wrong one.
ON THE RECORD'Two versions of the same play' asserts a single actor. The draft's own unknowns section says we do not know that. A finding should not contradict its own unknowns; if the desk keeps the phrase, it should delete the unknown and own the inference in public.
ON THE RECORDDropping 'hurricanes' from a list the FTC put first, while keeping five foreign place names, materially changes how relevant this alert looks to a domestic donor. I do not think that was intentional, but I think it is a real distortion of the source.
ON THE RECORDOmitting ReportFraud.ftc.gov from a fraud alert built entirely on an FTC page that ends by telling readers to report charity imposters there is the most consequential defect in this draft. Everything else is polish.
ON THE RECORDThe option lists are junk — 'ATTENTION: SCAMMEDS', 'Forward this to the disasters in your life', and 'giving'/'month' as candidate target groups. share_directive came out null, which is the right answer, but I do not credit the generator for it and I want the failure on the record.
The sources
Official sourceWhen donating, support those in need, not a scammer2026-08-27
The FTC states that scammers create fake charities that seem official or impersonate well-known charity organizations after disasters, and warns against donating to anyone who says you can only pay by cash, gift card, wire transfer or cryptocurrency.
Authority: official. Retrieved 2026-08-27.
Limitation: A prevention advisory only: it gives no case numbers, loss totals, impersonated organisation names, or geographic breakdown of who was approached.
Open the original source →Other checks
Published under standing founder pass (A9) — every claim source-mapped by the machine.
▼ What the machine checked
- ✓ Not a community submission.
- ✓ No entity is named.
- ✓ All 4 material sentence(s) map to FTC.
- ✗ anthropic returned "overstated"; google returned "overstated"; groq returned "overstated" — published on the receipt, not blocking (A9 amendment).
- ✓ No audience band is set.
No human affirmed these. They were verified by the classifier described in Amendment A9, on 2026-08-28.
Something wrong here? Tell us and we'll correct it — corrections are published, not quietly edited.
Phishy? Send it → sharelivefraud.com/squire-it
Not affiliated with any government agency, credit bureau, bank, platform, or law-enforcement agency. Informational only — not legal or financial advice.
Naming a source is not an endorsement, and being named here is not an accusation against any company.
Powered by SquireIt™
Verify this receipt at squireit.com