OpenAI has released its 2026 malicious AI report, published February 25, 2026, detailing how it detects and prevents hostile uses of artificial intelligence. The OpenAI News publication highlights two years of insights into evolving threat actor tactics, emphasizing the blended nature of AI abuse.
The report underscores that threat actors rarely rely on AI in isolation. Instead, they integrate AI models with conventional tools such as websites and social media accounts. This multi-platform approach, as illustrated by a case study on a Chinese influence operator, often involves employing different AI models throughout their operational workflow.
These findings are crucial for the industry and broader society to better identify and mitigate emerging threats. Understanding how OpenAI AI threat actors leverage multiple models, including advancements like GPT-5.2 and Sora 2, provides essential intelligence for developing more robust defense mechanisms.