OpenAI Shuts Down Russian, Iranian ChatGPT Influence Ops; Nearly 100 Fake Articles Planted Globally

OpenAI shut down two state-sponsored influence operations on October 8 using ChatGPT to run fake journalist personas and a front think-tank. The Iranian operation deployed seven fake Western journalist profiles that published nearly 100 articles across a dozen international outlets from July 2025 through October 2026; the Russian operation co-opted unwitting Latin American staff to run the Social Research Center, a front entity targeting Ukraine's reputation in the region.

The Iranian operation's 15-month runway — from July 2025 through October 2026 — before detection suggests OpenAI's content-moderation stack flags individual articles but not coordinated byline networks; the operation only accelerated after March's war began because the geopolitical urgency masked the pattern. The Russian operation's use of co-opted Latin American staff is the operational tell: it's cheaper than running fake personas at scale, avoids the linguistic fingerprints that sink Iranian fake-journalist schemes, and distributes enforcement burden across platforms that have no shared detection framework. If this model works in Latin America, expect similar front-staffing in Eastern Europe and Southeast Asia within 12 months.

WHY IT MATTERS

Both operations escalated after the March 2026 Iran war began — the Iranian network's publication frequency jumped, and the Russian front expanded its Latin America footprint.

This confirms state actors are weaponizing commodity AI APIs to run persistent disinformation at scale, bypassing traditional media gatekeeping and exploiting OpenAI's own moderation at the account level. The Iranian operation's seven personas on X and Instagram used platform-native transparency tools to mask their origin (spoofed as US-based writers; actually Iran-based), which means platform detection is lagging behind operational sophistication. Watch whether OpenAI, Meta, and X publish joint standards for detecting AI-generated influence accounts, or whether each platform independently tightens API-layer controls on account clustering and content velocity.

WHAT THIS DOESN’T TELL US

Did OpenAI detect these operations through user-behavior anomalies (account clustering, API call patterns) or through manual investigation of content patterns? If the former, how many similar clusters are still active and undetected?

Fault Lines

Strategic intelligence, synthesized daily — with a public track record. Every call graded against what actually happened.

Front page → Get the weekly brief →