Attackers prompted Gemini over 100,000 times while trying to clone it, Google says (arstechnica.com)

Google revealed that attackers have attempted to clone its Gemini AI chatbot, prompting it over 100,000 times to extract data for a potential cheaper imitation. This activity, described as "model extraction," is considered a form of intellectual property theft by Google. While the company presents itself as both a victim and proactive defender of its technology, the practice raises ethical questions, especially given that Gemini itself was trained on web-scraped data without explicit permissions. Such cloning efforts are reportedly driven by private companies and researchers seeking competitive advantages. The practice of distillation, where a new model is trained on the outputs of an existing one, is common in the AI/ML field. This underscores the ongoing tension between model developers and those who might seek to replicate their work without the associated resource investment. Google’s warning about these threats serves as a cautionary tale for the AI community, highlighting the need for stronger protections and ethical considerations in the rapidly evolving landscape of AI technologies.