OpenAI, the creator of ChatGPT, has revealed in a recent report that it has successfully blocked over 20 malicious operations and dishonest networks globally in 2024. These operations varied in scope, purpose, and target, ranging from malware creation to the generation of fake media accounts, fabricated biographies, and deceptive website articles.
OpenAI’s analysis of these thwarted activities offers valuable insights into the evolving landscape of AI misuse. While threat actors continue to explore and experiment with OpenAI’s models, the report indicates no significant breakthroughs in their ability to develop novel malware or cultivate substantial online followings. This finding holds particular significance in a year marked by elections in several countries, including the United States, Rwanda, India, and the European Union. In one instance, OpenAI banned multiple accounts in early July that were disseminating election-related comments about Rwanda across various X (formerly Twitter) accounts. This highlights OpenAI’s proactive approach to mitigating potential misuse during sensitive political periods.
Further demonstrating its commitment to security, OpenAI disrupted a China-based threat actor dubbed “SweetSpecter” that targeted OpenAI employees with spear-phishing attacks aimed at both corporate and personal email addresses. Additionally, OpenAI collaborated with Microsoft following the latter’s exposure of domains linked to the Iranian covert influence operation known as “STORM-2035.” OpenAI subsequently investigated, disrupted, and reported associated activity on ChatGPT stemming from this operation.
Importantly, OpenAI observed limited engagement with the social media posts generated by these malicious actors. The posts garnered minimal comments, likes, or shares, suggesting that their influence campaigns were largely ineffective. OpenAI emphasizes its ongoing commitment to anticipating and counteracting the malicious use of advanced AI models, pledging to take the necessary steps to prevent harmful activities.
