OpenAI Disrupts Russian and Iranian Influence Operations Using ChatGPT
OpenAI’s safety team reports disrupting propaganda efforts by Russian and Iranian actors that exploited ChatGPT for fake identities and disinformation.
Key points
- OpenAI disrupted influence operations linked to Russian and Iranian actors.
- ChatGPT was used to support a Latin American propaganda campaign.
- The service was also used to boost Iranian fake journalist identities.
OpenAI announced that its safety team has disrupted coordinated influence operations conducted by Russian and Iranian actors who exploited the ChatGPT platform. According to a report by The Record, these efforts involved using the AI assistant to generate content for propaganda campaigns and to create deceptive online personas.
The company’s safety team identified two distinct but related abuse patterns. First, ChatGPT was utilized to support a propaganda operation targeting Latin America, which OpenAI attributes to Russian actors. Second, the service was used to amplify a set of fake journalist identities linked to Iranian actors. In both instances, the safety team intervened to disrupt the activities, removing the offending content and accounts associated with the inauthentic behavior.
Context and Background
The disclosure highlights the ongoing challenge of preventing state-aligned actors from leveraging generative AI tools for disinformation. According to The Record, OpenAI’s safety team specifically cited the use of ChatGPT to boost the credibility and reach of these fake identities and propaganda narratives. The intervention demonstrates the company’s active monitoring of its platform for coordinated inauthentic behavior, particularly when linked to foreign influence operations.
By identifying and disrupting these specific campaigns, OpenAI aims to reduce the effectiveness of automated or AI-assisted disinformation. The report does not specify the exact dates of the operations or the specific political events targeted in Latin America, nor does it detail the technical methods used by the actors to bypass initial safety filters. However, the confirmation that the safety team successfully identified and acted against these threats provides insight into the current state of AI security defenses against sophisticated, state-linked adversaries.
Who Is Affected
The primary targets of these influence operations were audiences in Latin America, where the Russian-linked propaganda campaign was directed. Additionally, the broader information ecosystem was affected by the presence of Iranian-linked fake journalist identities, which sought to manipulate public perception through deceptive authorship.
Security professionals and platform operators monitoring for disinformation should note that generative AI tools are being actively used to scale the production of fake personas and propaganda content. The disruption of these specific operations indicates that such activity is detectable and actionable, but it also underscores the persistent risk of AI-assisted influence campaigns. Users and organizations relying on online information sources may have been exposed to content generated or amplified by these inauthentic accounts before the safety team’s intervention.
What should you do?
As generative AI tools become more prevalent, the risk of AI-assisted disinformation and influence operations continues to evolve. Security teams and individual users should remain vigilant for signs of coordinated inauthentic behavior, such as sudden spikes in engagement from new accounts or content that exhibits consistent stylistic patterns indicative of AI generation. Monitoring for unusual activity and verifying the authenticity of sources are critical steps in mitigating the impact of such campaigns.
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
- Monitor social media and news feeds for sudden surges in engagement from newly created accounts, which may indicate coordinated inauthentic behavior.
- Verify the authenticity of online sources and journalists by cross-referencing information with established, reputable news outlets and checking for consistent digital footprints.
- Review platform security settings and enable additional authentication measures to reduce the risk of account takeover or impersonation.
- Stay informed about emerging threats involving generative AI by following updates from cybersecurity vendors and official safety team disclosures.
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Which countries were linked to the influence operations disrupted by OpenAI?
OpenAI’s safety team linked the operations to Russian and Iranian actors.
What specific regions or topics were targeted by these campaigns?
The Russian-linked campaign targeted Latin America with propaganda, while the Iranian-linked effort involved boosting fake journalist identities.
Did OpenAI confirm that a patch or update was released to prevent this?
The source material states that the safety team disrupted the operations, but it does not mention the release of a specific software patch or update.
