Western media faces top-rated AI influence ops from Iran, Russia per OpenAI's first high-severity alert

OpenAI has identified coordinated influence campaigns linked to Iran and Russia as high-severity threats, assigning them the highest ratings the company has ever used for such activities. According to CyberScoop, these operations represent the first time OpenAI has formally reported what it considers to be high-impact influence campaigns.

The disclosure highlights a shift in how the AI developer categorizes state-sponsored disinformation, moving beyond standard threat monitoring to explicit severity grading.

What specific actions did OpenAI take?

OpenAI assigned severity scores of 4 and 5 out of 6 to the two distinct campaigns. These ratings indicate a significant level of threat and impact, distinguishing them from lower-priority anomalies. By publicizing these scores, OpenAI is establishing a new benchmark for evaluating the danger posed by AI-driven influence operations.

This is the first instance where the company has reported influence campaigns that it deems high-impact. The decision to rate and disclose these specific operations suggests a tightening of OpenAI’s reporting standards regarding state-actor threats.

Who was behind these campaigns?

The influence operations were attributed to Iran and Russia. According to CyberScoop, these state actors utilized AI-generated content to push their narratives. The campaigns were not random but were structured efforts to alter public perception and media coverage in Western countries.

The involvement of these two nations points to a sophisticated use of AI tools for geopolitical influence. Rather than relying solely on human-operated accounts, the campaigns leveraged automated or AI-assisted methods to scale their reach and mimic legitimate sources.

How did the campaigns operate?

The operations employed AI-generated personas, including fake journalists and think tank representatives. These fabricated entities were designed to appear as credible sources of information within Western media ecosystems. By posing as experts or reporters, the campaigns aimed to inject disinformation into mainstream discourse under the guise of authority.

This method of infiltration allows state actors to bypass traditional skepticism directed at obvious propaganda. The use of think tank and journalist personas adds a layer of legitimacy that makes the disinformation harder to detect and counter for average consumers and editors alike.

What does this mean for organizations?

For software developers and security teams, this report serves as an early warning about the evolving tactics of state-sponsored influence. The high severity ratings from OpenAI suggest that AI-generated content is becoming a primary vector for geopolitical interference. Organizations should remain vigilant about the sources of information they encounter, particularly when dealing with AI-generated text or media.

No specific technical fixes or patches have been confirmed for these influence campaigns, as the threat lies in the content and narrative rather than a software vulnerability. However, the disclosure underscores the need for robust detection mechanisms for AI-generated personas and coordinated inauthentic behavior.

What to do and how to stay safe: OpenAI

• Data classification must precede any permission to use generative tools.
• Automated agents require stricter isolation than simple chat interfaces.
• Audit trails must capture both prompts and responses for forensic analysis.

AI policies must balance innovation with strict data controls and human oversight. Start by classifying your data and restricting access to public models.

Step-by-step guide: AI Usage Policy Checklist for Secure Enterprise Deployment

General security guidance from the Sandbox Desk newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

How severe did OpenAI rate these campaigns?

OpenAI rated the campaigns 4 and 5 out of 6 for severity, the highest ratings it has assigned to influence operations.

Which countries were responsible for the campaigns?

The campaigns were attributed to Iran and Russia, according to CyberScoop.

Did OpenAI provide a fix for this threat?

No fix has been confirmed yet, as this is an influence campaign rather than a software vulnerability.