OpenAI bans Russian and Iranian 'false front' influence operations—its first Category 5 case, with AI mostly used to write reports to bosses

Why it matters

【Key points】

OpenAI's Oct 8 threat report says it banned two foreign influence operations that paired AI with traditional tradecraft. Russia-origin "Dark Clark" ran a Latin American "Social Research Center" through a fake persona, "Mia Clark," using unwitting local staff to produce content aimed at undermining Ukraine and meddling in politics in Argentina, Bolivia and elsewhere. The operators said they faked an email from Lima's education directorate getting schools to honor Stepan Bandera, then pushed stories into Peruvian and Polish media, inflaming Polish–Ukrainian tensions; they also faked audio and documents in Ecuador and Bolivia. OpenAI found matching coverage, fact checks and official denials.

Iran-origin "Bogus Bylines" used seven fake journalist personas to pitch long-form articles to small and mid-sized outlets and mass-produced comments on the US–Iran war. On the 1–6 Breakout Scale, OpenAI rated the Russian operation Category 5, a first among the 30 it has exposed in two and a half years; Iran's reached Category 4. Both used ChatGPT most for internal reports, with signs of inflated results.

【Verifiable】

OpenAI report: https://openai.com/index/disrupting-ai-enabled-false-front-operations

【Our read】

1. AI added efficiency, not new tactics. Fake journalists and think tanks are decade-old playbooks; what got them into mainstream outlets was editors' and schools' trust in pitches and letters.

2. The top use was padding reports to superiors. Influence operations' internal KPIs are being inflated by AI too, so claimed impact deserves a discount.

3. This visibility exists because OpenAI can read operators' prompts. Model providers are becoming intelligence sources, but the scoring and attribution are OpenAI's own; outsiders can verify only the publicly searchable parts.