This study provides empirical evidence of LLM-generated content in real-world disinformation, moving beyond theoretical concerns to documented presence.

• Found increasing prevalence of LLM-generated text in multilingual disinformation campaigns
• Identified unique linguistic patterns that distinguish AI-generated content across languages
• Revealed that current detection methods have significant limitations in real-world contexts
• Demonstrated concrete security risks in specific "longtail" contexts previously considered low-risk

These findings challenge the narrative that LLM misuse concerns are overblown, highlighting actionable security vulnerabilities that require attention from both technology providers and security professionals.

Beyond speculation: Measuring the growing presence of LLM-generated texts in multilingual disinformation