OpenAI Bans Russian and Iranian Influence Networks for Using ChatGPT
A new report details how two covert operations used generative AI to accelerate traditional disinformation campaigns targeting Western media outlets.
Key points
- OpenAI banned two influence networks originating from Russia and Iran after tracing their use of ChatGPT to create false-front entities.
- The Russian network achieved a Category 5 rating on the IO Breakout Scale, marking the first operation at this level disrupted since tracking began.
- Both groups placed geopolitical content in mainstream media outlets rather than relying solely on social media channels, according to the report.
OpenAI has banned two covert influence networks, one linked to Russia and another to Iran, after determining they used ChatGPT to accelerate traditional disinformation tactics. The decision follows a new threat report from the company that traced how these groups built false-front entities to push conflict-related messaging into target audiences.
The central finding of the OpenAI report is that the actors did not invent new methods but instead used generative AI to speed up old playbooks. Both networks were identified for constructing personas and front organizations designed to make their geopolitical content appear legitimate to Western readers and journalists.
Return to Pre-AI Playbooks
The Iranian network’s personas resemble the style of "Alice Donovan," a front for Russian military intelligence whose articles appeared in Western outlets in 2016. Similarly, the Russian network’s use of unwitting staff mirrors the operations of "PeaceData," a fake outlet run in 2020 by individuals linked to the Internet Research Agency. These comparisons highlight a continuity in method despite the new tools.
According to The Cyber Express, the primary difference in these recent operations lies in their reach and integration into traditional media ecosystems. Both groups successfully placed content in mainstream outlets rather than relying only on social media distribution, allowing their narratives to gain credibility through established journalistic channels.
Severity and Reach
On the IO Breakout Scale, which rates operations from one to six, the Russian network was assessed at Category 5. This assessment makes it the first operation at this severity level to be disrupted since the reporting began, indicating a significant escalation in operational sophistication. The Iranian network reached Category 4, demonstrating a substantial capability to penetrate media landscapes.
The high ratings reflect the success of these groups in placing content in mainstream outlets rather than relying only on social media. The report frames the activity as a return to pre-AI playbooks, noting that the difference lies in reach compared to previous operations.
What should you do?
Security and technology professionals should monitor for sudden spikes in content from new or obscure journalistic sources that align with specific geopolitical narratives. Organizations relying on user-generated content or external submissions should implement stricter verification processes for author identities and historical publishing records.
Practical Mitigation Steps
Background: Generative AI
Generative AI creates content from prompts. Agentic AI performs actions using tools. The former outputs text or code; the latter changes system states. Understanding this distinction is critical for security architecture, as agentic systems introduce execution risks that pure generation models do not possess.
Read the full guide: Generative AI vs Agentic AI: Architectural Differences and Security Risks
What to do and how to stay safe: OpenAI
- Verify the provenance of all contributed content by cross-referencing author identities against known threat actor profiles and previous publishing histories.
- Implement technical checks for AI-generated text patterns in submissions from new or unverified sources to identify potential automated influence campaigns.
- Require multi-factor identification for any new contributor accounts, particularly those submitting content related to sensitive geopolitical or conflict-related topics.
- Train editorial and security teams to recognize the characteristics of false-front entities, such as inconsistent biographical details or sudden, high-volume publishing bursts.
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
What specific tools did the Russian and Iranian networks use?
According to the OpenAI threat report, both networks used ChatGPT to speed up their operations and build false-front entities for their disinformation campaigns.
How severe was the Russian network's operation?
The Russian network was assessed at Category 5 on the IO Breakout Scale, making it the first operation at this level disrupted since the reporting began.
Did these networks use new tactics?
No, the OpenAI report states the actors used ChatGPT to speed up old tactics rather than inventing new ones, resembling pre-AI playbooks from previous years.
