OpenAI Disrupting Malicious Uses of AI Report June 2025

OpenAI is utilizing AI-driven investigative tools to detect and disrupt malicious activities such as covert influence operations, cyber espionage, and social engineering. This approach serves as a force multiplier for expert teams, enabling the identification and exposure of abusive activities more efficiently.

Strategic Framework for AI Safety and Governance

OpenAI operates under a framework that prioritizes protecting people from actual harms through common-sense rules and the development of democratic AI. This strategy focuses on preventing AI tools from being used by authoritarian regimes to consolidate power, control citizens, or coerce other states.

Key areas of focus for preventing AI abuse include:

  • Covert influence operations (IO)
  • Child exploitation
  • Scams and spam
  • Malicious cyber activity

Recent Disruptions and Case Studies

In the three months preceding the June 2025 report, OpenAI's investigative teams used AI to detect and expose several categories of abusive activity. By integrating AI into their defense mechanisms, the organization has successfully disrupted the following:

  • Covert Influence Operations: Identifying deceptive campaigns designed to manipulate public opinion.
  • Cyber Espionage: Detecting attempts to use AI for unauthorized data access or intelligence gathering.
  • Social Engineering: Preventing attacks that rely on psychological manipulation to deceive users.
  • Deceptive Employment Schemes: Exposing fraudulent job offers used as vectors for scams or data theft.
  • General Scams: Disrupting various AI-enabled fraudulent activities.

Alignment with U.S. AI Action Plan

These efforts align with OpenAI's March submission to the Office of Science and Technology Policy’s U.S. AI Action Plan. The organization maintains that ensuring AI benefits humanity requires a dual approach: implementing rules to prevent harm and deploying AI tools to empower those defending against systemic abuses.

Sources