Katarzyna Petru
Google has published a new security report detailing the growing attempts to copy advanced artificial intelligence models. The targets of these attacks include Gemini, the company's flagship AI model. According to the document, attackers used over 100,000 queries to analyze and recreate the system's functionality. Google claims that these actions are part of a broader wave of attempts to steal intellectual property related to AI. The attacks are said to originate from countries such as North Korea, Russia, and China. However, the company emphasizes that the threat primarily concerns model creators rather than ordinary users.
What is "model extraction"?
Google describes these actions as so-called extraction attacks, which are systematic examinations of a model using legitimate access. In practice, this means mass-generating queries to understand the mechanisms of AI operation and using that knowledge to create a competitive system. Such operations may lead to the creation of models in other languages or to replicating functionality without access to the original training data.
Arms Race in the AI World
Experts from the Google Threat Intelligence Group warn that the company may be merely the first example of a larger problem. In recent years, competition in the AI sector has significantly accelerated, and tensions between tech companies are rising. New advanced models from Asian companies have emerged on the market, increasingly competing boldly with leaders from the USA. In the background, disputes are also underway regarding the use of others' technology to train their own systems.
The Google report shows that as artificial intelligence evolves, the risk of technology theft also increases. Attacks involving mass analysis of models may become one of the main threats to companies developing AI. For users, this means one thing – the race for dominance in the world of AI is just getting started.
source: cnet.com
