Latest News: Key Developments from the Region – Impakti

News – Over 100,000 Prompts Used in Attempts to Copy Gemini

More than 100,000 requests (“prompts”) have been identified as part of efforts to extract capabilities from Google’s artificial intelligence models, according to a report published by the Google Threat Intelligence Group. According to Google, these activities are associated with so-called “distillation attacks,” in which various actors attempt to replicate the functionality of advanced models through repeated and systematic queries. The report, titled “Distillation, Experimentation, and Continued Integration of AI for Adversarial Use,” examines how malicious actors are experimenting with and integrating artificial intelligence into their operations. Google states that some campaigns involved over 100,000 prompts directed at the company’s models, with the aim of extracting information about their behavior and functional structure. These efforts are described as model extraction attacks , where users exploit legitimate access to a generative model to collect responses at scale, attempting to build alternative models with similar capabilities. Google emphasizes that this practice violat…