# AI Runs Fraud and Social Engineering at Platform Scale

Single small operators increasingly use frontier models to run thousands of synthetic personas and millions of targeted messages, so expect recurring lab threat reports quantifying mass-persona fraud, account-level enforcement as the main countermeasure, and pressure on consumer platforms and payment rails to detect machine-run relationships rather than machine-written text.

- Conviction: 38 / 100 (forming)
- Horizon: Emerging (watchlist)
- Tracking since: 2026-09-14T00:00:00.000Z
- Last updated: 2026-09-14T14:04:09.672Z
- Canonical: https://polylog.news/ai/trends/ai-industrialized-fraud-operations
- Publisher: Polylog
- Affected regions: Global

## Recent evidence

- [confirms] Anthropic Says It Disrupted Claude Misuse Across Seven Harm Categories in Nine Months (2026-09-14): Anthropic reported one disrupted operation that used Claude to run more than 4,700 dating app personas, sending 2.36 million messages to at least 25,000 people in two weeks. The economics of romance and affinity fraud shift when persona count and message volume stop scaling with operator headcount.