Shypd.aiThe largest open index of AI tools
Attackers prompted Gemini over 100,000 times while trying to clone it, Google says. This highlights the use of 'distillation' techniques to mimic large AI models at a fraction of the development cost.
Trending
About
This article from Ars Technica reports on Google's findings that "commercially motivated" actors have prompted its Gemini AI chatbot over 100,000 times in an attempt to clone its knowledge. This practice, known as "model extraction" or "distillation," involves feeding an existing AI model thousands of prompts, collecting the responses, and then using these input-output pairs to train a smaller, cheaper copycat model. Google considers this intellectual property theft, though the article notes the irony given Google's own LLM was built from scraped internet data. The technique allows copycats to replicate many of a parent model's capabilities without access to its code or original training data, offering a significant shortcut in AI development.
Capabilities
Pricing & Plans
unknown
Not publicly disclosed
FAQs
Trending
