From Artificial Intelligence News: 2024-05-31 13:54:24

In the last three months, OpenAI has thwarted five covert influence operations intending to exploit AI models for deceptive online activities. No significant increase in audience engagement was observed due to OpenAI’s interventions.

OpenAI’s commitment to safety and AI model design has been key in disrupting threat actors’ attempts to create deceptive content. AI tools have also boosted the efficiency of OpenAI’s investigations into covert influence operations.

Cooperation from distribution platforms and the open-source community has greatly aided in combating influence operations. OpenAI is sharing its findings to encourage information sharing and best practices among stakeholders in the broader community.

OpenAI has disrupted several covert influence operations utilizing its models for various tasks such as generating comments, creating fake profiles, and translating text. Operations targeted issues like political conflicts, elections, and government criticism across different platforms.

Investigations into influence operations revealed trends including AI-enhanced content generation, mixing traditional and AI formats, fake engagement tactics, and productivity gains through AI tools. Defensive measures include friction on threat actors, AI-enhanced investigations, and the importance of effective distribution.

Despite the challenges of detecting and disrupting covert influence operations, OpenAI remains committed to developing responsible AI models that actively intervene against malicious use. The company emphasizes the importance of industry sharing and the human element in threat actor errors.



Read more at Artificial Intelligence News: OpenAI disrupts five covert influence operations