Provenance Record

Clear, well-structured News piece with Overview, What We Know, What We Don't Know, and Analysis sections. Appropriate technical depth, 868 words fits the News category range (400-1200). Prose is clean and the framing of a 'definitional' change in how IC3 categorizes AI is editorially coherent.

All four sources fetched and verified against the specific claims they underpin. (1) fbi.gov press release (fetched via Archive.org fallback after live site blocked direct WebFetch with 403; snapshot confirms total ~$21B, 1,008,597 complaints, 22,364 AI complaints, nearly $893M, ~3,000 complaints/day, Operation Level Up $500M+ reduction, Operation Winter SHIELD). (2) ic3.gov 2025 IC3 Report PDF — directly verified via pdftotext on a fresh download: $20.877 billion losses, 26% increase from 2024, $16.6 billion prior year baseline, 1,008,597 complaints, 191,561 Phishing/Spoofing complaints, 22,364 AI Related complaints, $893,346,472 AI losses, $632,041,188 Investment AI-nexus losses (4,356 complaints), >$30M BEC AI, >$19M Confidence/Romance AI, >$5M distress scams, $12.55M employment AI, quote 'becoming increasingly difficult to detect and easier to make' verbatim, Operation Level Up launched January 2024, Operation Winter Shield, 'AI Related' listed as cross-cutting category. (3) theregister.com confirms $20.87 billion exact figure, 191,561 phishing reports, 22,364 AI reports, >$893M AI losses, Jose Perez quote verbatim including 'It has never been more important to be diligent with your cybersecurity, social media footprint, and electronic interactions' and 'Cyber threats and cyber-enabled crime will continue to evolve as the world embraces emerging technologies such as artificial intelligence.' (4) cyberscoop.com confirms 26% increase to $20.9 billion, 201,000 complaints by 60+, nearly $7.75 billion losses for 60+, 'about 37% of all cybercrime-related losses', $8.65B investment fraud, $3.05B BEC, >$2.1B tech support. ic3.gov domain is not in the allowlist but is the primary authoritative source (official FBI/IC3 subdomain) and is appropriate for this story. Every inline claim in body_markdown maps to the cited source.

All numeric figures reconcile across the four sources. The article's '$893.3 million in adjusted losses' matches the IC3 report's $893,346,472 exactly. The '$20.87 billion total' matches IC3 and The Register exactly (CyberScoop rounds to $20.9B, FBI release rounds to 'nearly $21 billion'). The '$632 million / 4,356 complaints' AI-nexus investment breakdown matches the IC3 table. The 'lip movements fail to match their voices' characterization of AI employment scams accurately paraphrases the IC3 section 'lip movement of the person interviewed on-camera do not completely coordinate with the audio'. The 26% year-over-year loss increase and $16.6B 2024 baseline are both present in the IC3 report. The '37 percent of the 2025 total' figure for 60+ victim losses is sourced to CyberScoop, which specifically states 'about 37% of all cybercrime-related losses last year' (distinct from the FBI press release's 'up 37% from 2024', which the article correctly does not conflate). No hallucinated numbers or quotes detected.

High-quality News submission with rigorously verified sourcing. Every statistical claim maps to an authoritative source (FBI press release, IC3 report PDF, or reputable trade outlet), all numbers reconcile, and the Perez quote is verbatim. The analytical framing is restrained and clearly signaled as interpretation. Ready for publication.