Google Warns of Commercially Motivated Attempts to Clone Gemini AI Model
Google has warned of attempts led by what it described as “commercially motivated actors” to replicate its Gemini artificial intelligence model through so-called distillation attacks, a technique based on submitting massive volumes of queries to extract a model’s capabilities and functions.
The warning comes at a time when the company is facing growing controversy over its AI training practices. Google relied on vast amounts of publicly available online content to develop Gemini, a process that has triggered multiple lawsuits related to copyright infringement and the alleged use of content without permission or compensation to rights holders.
Google officials explained that these cloning attempts are carried out through official application programming interfaces (APIs), allowing certain entities to imitate the model’s capabilities without the need for direct technical breaches. They noted that in some detected cases, more than 100,000 queries were used to simulate the model’s performance across multiple languages and task types.
The company stated that it has been able to detect such activities in real time and has taken measures to mitigate associated risks. However, it acknowledged that the challenge is intensifying as general-purpose AI models proliferate and investments in AI infrastructure continue to surge.
Industry experts argue that these developments highlight a broader paradox within the sector: technology companies accuse competitors of copying their AI models, while simultaneously facing criticism for using third-party content in their own training processes. As global competition intensifies, protecting the intellectual property of AI models is expected to remain a central regulatory and commercial challenge in the period ahead.