What we found
The FBI has reissued a public warning about widespread schemes in which scammers impersonate US and foreign law enforcement or government officials to extort money or personally identifiable information from victims. Between January 2025 and July 2026, the FBI's Internet Crime Complaint Center received nearly 61,000 complaints of law enforcement or government impersonation scams, with losses totaling more than $1.6 billion. Contact comes primarily through unsolicited phone calls, with text messages and emails also used, and the callers spoof real agency phone numbers, email addresses, employee names and credentials — including those of IC3 itself. The FBI describes an urgent, aggressive script: the caller refuses to speak to anyone but the target, keeps the person on the phone for the duration of the scam, and urges them not to tell family, friends, their bank or law enforcement. Most government impersonation complaints involve the caller accusing the person of a crime and threatening arrest, prosecution or imprisonment unless a payment is made, while other pretenses include a missed jury duty or court date — 6,833 complaints and nearly $36 million in losses — and expiring passports, driver's licenses or professional licenses, with 496 complaints and more than $348,000 in losses. Medical practitioners are targeted with claims that their license is expiring or was used in a crime, accounting for 3,322 complaints and more than $37 million in losses over the same period. Scammers also target Americans from different ethnic communities, foreign nationals and international students in the United States, posing as foreign law enforcement or US-based foreign diplomatic officials and threatening extradition or cancellation of a home country passport — 1,809 complaints and more than $140 million in losses — sometimes wearing fake police uniforms and using mock sets of official government facilities. The FBI states the payment demands include prepaid cards, couriers, bank wires, cryptocurrency, or cash inserted into cryptocurrency kiosks, and that real law enforcement and government officials will never phone or text the public to demand payment or sensitive information. The FBI also says scammers use new technologies such as artificial intelligence to appear as officials on video calls, adding the appearance of legitimacy. One action, every time: hang up, then look up the agency's number yourself from an official public source and call back — never use a number, link or email address the caller gave you. If money has already moved, stop all contact with the caller, tell your bank straight away, file a report with local law enforcement and file a complaint at ic3.gov, keeping the messages and financial records. The advisory does not say how many of these complaints involved AI-generated video calls, and it does not report victim ages, incomes or locations. We harvested no press or community reports alongside this advisory, so the figures here rest on one official source and its own complaint data. LIVEFRAUD ALERTS is independent and not affiliated with the FBI, IC3 or any government agency; this alert summarises a public advisory and adds our own guidance.
- The sole source is an official FBI/IC3 public service announcement with a publication date and complaint counts broken out by scam type.
- Every claim sentence maps directly to explicit statements or figures in that advisory, with no inference beyond its wording.
- Confidence is in the advisory's own reporting only — complaint data is self-reported by victims and no second source was harvested to corroborate it.
- Reviewed by 3 models, 2 from independent houses.
THE RULE
High-pressure calls from spoofed government numbers demanding prepaid cards, wires or crypto — more than $1.6 billion in reported losses in 19 months.
What we don’t know
- How often the AI-assisted video call tactic appears among the complaints.
- Which specific agencies or brands were spoofed beyond IC3, and what numbers or domains were used.
- Whether any of the reported funds were recovered.
- The ages, incomes or locations of the people who filed complaints.
The bench — who voted
3 INDEPENDENT AI MODELS REVIEWED THIS. ALL 3: HIGH RISK.
The card names a count. Here are the seats behind it, with what each one said.
ANTHROPICclaude-opus-5 · HIGH
The single evidence row supports only the aggregate figures (~61,000 complaints, >$1.6bn, Jan 2025–Jul 2026, spoofed numbers, arrest threats, payment rails). The finding adds five distinct sub-statistics that appear nowhere in the evidence: jury duty/court 6,833 complaints and ~$36m; expiring passports/licences 496 complaints and >$348,000; medical practitioners 3,322 complaints and >$37m; foreign-national/diplomatic impersonation 1,809 complaints and >$140m. Each of these is a precise, checkable number resting on nothing the panel was shown.
DISSENT, PRESERVEDThe headline CLAIM as written — impersonation of law enforcement/government officials, ~61,000 complaints, >$1.6bn, Jan 2025–Jul 2026 — is fully carried by the evidence row. My 'overstated' verdict attaches to the elaborated finding, not to the core numbers, and readers should not discount the alert on that basis.
DISSENT, PRESERVEDThe limitation that these are victim-reported complaints, not a count of all incidents, belongs in the published text and is currently absent. Under-reporting means the true scale is likely larger, not smaller, so this caveat cuts against complacency rather than against the alert.
DISSENT, PRESERVEDThe operational advice — hang up, independently look up the agency number, never use a number or link the caller supplied — is correct and is the single most protective instruction here. I would keep it even if the panel strips the disputed statistics.
GROQopenai/gpt-oss-120b · HIGH
The finding adds numerous details (e.g., primary contact via phone, use of texts/emails, spoofed employee names, specific script language, sub‑category complaint counts, AI‑generated video calls, and step‑by‑step user guidance) that are not present in the single evidence row, which only reports aggregate complaint and loss figures and mentions spoofed numbers, arrest threats, and payment methods.
OPENROUTERopenrouter/free · HIGH
The evidence row supports the overall complaint and loss totals, spoofed agency numbers, arrest threats, and listed payment methods, but not the finding's detailed subcategory counts and losses, foreign-official scenarios, uniforms and mock facilities, or medical-practitioner targeting.
DISSENT, PRESERVEDThe core warning and headline totals are supported by the evidence row, but the finding contains many specific subcategory totals, tactics, scripts, and recommendations that the supplied evidence does not establish.
Reviewed by 3 independent models; all judged the finding to go beyond the evidence.
▼ Protocol & challenge record
Objection (medium): DIRECTIVE OPTION NOT IN EVIDENCE: 'Send this to any official you know.' The advisory describes officials as the entities being *impersonated*, not as a targeted victim group. Offering this directive implies officials are being targeted and breaches the same §11 Rule 2 logic the pipeline used to drop three other directives. It should be removed from directive_options, not left selectable. (The other three options — practitioners x2, international students — are squarely evidenced by the 3,322 and 1,809 complaint paragraphs and are fine.)
Not resolved — preserved on the record.
Objection (medium): UNSOURCED ARITHMETIC IN THE RISK LINE: 'more than $1.6 billion in reported losses in 19 months.' The advisory never states a duration. 'Between January 2025 and July 2026' is ambiguous as to endpoint inclusivity — inclusive it is 19 months, exclusive it is 17-18. The risk line is the most-read sentence in the alert and it is the only place a number appears that is not in the source. Either drop '19 months' or write 'over roughly a year and a half'.
Not resolved — preserved on the record.
Objection (medium): PROTECTIVE ADVICE GAP ON THE HIGHEST-LOSS-PER-COMPLAINT VECTOR. The foreign-official/diplomatic scam is the worst ratio in the advisory (1,809 complaints, >$140M — roughly $77k per complaint, versus ~$26k overall). The FBI explicitly warns these 'can be challenging to verify due to the threat involving unknown officials overseas.' The draft reproduces the complaint figures but silently drops that verification caveat, while offering 'hang up and look up the agency's number yourself' as the one action every time. For an international student being told their home-country consulate is calling, the single prescribed action is precisely the one the FBI says is hard to execute. The caveat should be restored and the advice should add: contact your consulate or university international-student office through a number you already hold.
Not resolved — preserved on the record.
Objection (low): OMITTED HIGH-VALUE RED FLAG. The advisory's second protection bullet — 'Law enforcement and government officials will never request payment via prepaid cards, cryptocurrency, or courier' — is a cleaner, more actionable tripwire than most of what the draft kept, and it is dropped. The draft mentions those payment methods only as a description of scammer behaviour, not as a bright-line rule. Also dropped: the .com-vs-.gov lookalike-domain warning, which is the only part of the advisory addressing the email/text channel the draft itself says is in use.
Not resolved — preserved on the record.
Objection (low): CONJUNCTION SHIFT IN THE HEADLINE CLAIM. Source: 'extort money or personally identifiable information.' Draft claim: 'extorting money and personal information.' 'And' implies both are taken in a given incident. Restore 'or'.
Resolved: Trivial fix; the finding body already uses 'or' correctly ('extort money or personally identifiable information'), so only the top-level claim string needs the one-word change.
Objection (low): 'HIGH' CONFIDENCE ON A SINGLE SELF-REPORTED DATASET. The confidence_reasons are internally honest but the label is doing work the reasons undercut: one source, zero corroboration, victim-self-reported loss figures with no independent audit, no recovery data, and no denominator for how many complaints were verified as actual losses versus attempts. 'High' is defensible only as confidence that the FBI said this, not that the figures are accurate. If the rubric does not distinguish those two things, the label overstates. I would set it to medium or annotate it explicitly as 'high (attribution only)'.
Resolved: Partially resolved in-draft: the third confidence_reason and both limitation sentences state plainly that the data is victim-reported, uncorroborated and single-sourced, and the evidence row carries the same limitation. The disclosure is adequate; only the summary label remains overstated.
Objection (low): CLOSE-PARAPHRASE DENSITY. Long stretches of the finding track the PSA almost clause-for-clause ('urgent and aggressive', 'mock sets of official government facilities', 'Americans from different ethnic communities', 'adding the appearance of legitimacy'). It is accurate, but it is closer to reproduction than summary and inherits the source's vague phrasing ('Americans from different ethnic communities' — which communities? the advisory never says, and neither does the alert). Flagging as an editorial and transparency issue, not an accuracy one.
Resolved: No accuracy defect — every figure and characterisation checks out against the row, including 61,000/$1.6B, 6,833/~$36M, 496/>$348,000, 3,322/>$37M and 1,809/>$140M. Recording as style, not as a check failure.
Objection (low): CITATION HYGIENE UNVERIFIED. The harvested URL is given as 'ic3.gov/PSA/2026/PSA260917' without the www subdomain, and no alert number (IC3 PSAs normally carry one, e.g. I-0917xx-PSA) is captured, even though the draft correctly cites the superseded alert number I-030722-PSA from the body text. A reader who cannot resolve the link cannot check the figures — which matters more than usual here because this is a single-source alert. Capture the canonical URL and the current alert number.
Not resolved — preserved on the record.
Objection (low): PIPELINE ARTIFACT IN targeting_dropped. The three dropped directives are recorded as the bare tokens 'take', 'badge', 'check' — these are not group names and read as a parsing failure. The attached reason ('no harvested source describes this group') therefore cannot be audited: nobody can tell what was dropped or whether it was dropped correctly. This is a logging defect that should be fixed before publication, since the log is the only evidence the §11 screen ran properly.
Not resolved — preserved on the record.
Preserved dissent
ON THE RECORDI do not accept 'high' confidence on a one-source alert built entirely on unaudited, victim-self-reported complaint data with no recovery figures and no second source. The figures are accurate as a report of what the FBI published; they are not established as an accurate measure of the harm. The confidence label should distinguish those, and as written it does not.
ON THE RECORD'19 months' in the risk line is not in the source. It is the checker's own arithmetic on an ambiguous date range, placed in the single most-quoted sentence of the alert. Small, but it is exactly the kind of quiet number-invention this process exists to catch, and I object to it shipping.
ON THE RECORDDropping the FBI's own caveat that foreign-official threats 'can be challenging to verify' while keeping the $140 million figure from that same paragraph is the most substantive editorial failure in this draft. It keeps the fear and discards the honesty about how hard the fix is, for the group facing the largest average loss in the advisory. I would not sign off without it restored.
ON THE RECORD'Send this to any official you know' should never have survived the same screen that killed three other directives. That it did, alongside a targeting_dropped log containing the words 'take', 'badge' and 'check', tells me the §11 screen ran on garbage input and its clean-looking output should not be trusted on this run.
The sources
Official sourceScammers Impersonating Law Enforcement and Government Officials in Fraud Schemes2026-09-17
The FBI's IC3 received nearly 61,000 law enforcement and government impersonation complaints between January 2025 and July 2026, with losses over $1.6 billion, involving spoofed agency numbers, threats of arrest, and demands for prepaid cards, couriers, bank wires, cryptocurrency or cash into crypto kiosks.
Authority: official. Retrieved 2026-09-17.
Limitation: Figures are victim-reported complaints to IC3 rather than a count of all incidents; the advisory gives no demographic breakdown and does not quantify how often AI video calls were used.
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 9 material sentence(s) map to FBI/IC3.
- ✗ anthropic returned "overstated"; groq returned "overstated"; openrouter 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 2026-09-18.
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