Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
In new Threat Tracking Report Published Thursday, Google said the hackers were engaging in “distillation attacks,” including one case in which they used more than 100,000 AI requests to steal the company’s technology for its Gemini AI model.
Google said the attacks appear to be coming from adversaries in countries including North Korea, Russia and China, and that attempts to steal AI intellectual property and potentially replicate it in AI models in other languages are part of a broader set of AI- and malware-based attacks the company has seen.
The company defines these attempts as model extraction attacks, which, it says, “occur when an adversary uses legitimate access to systematically investigate a mature machine learning model to extract information used to train a new model.”
This could mean using artificial intelligence to bombard Gemini with thousands of prompts to replicate their typical abilities. Google indicated in the report that this does not pose a threat to its users, but rather to service providers and form creators, who may be vulnerable to their work being stolen and duplicated.
AI competition and AI theft
John Hultquist, senior analyst at Google Threat Intelligence Group, which prepared the report, He told NBC News Google may be one of the first companies to face these types of theft attempts, but there could be more. “We will be like the canary in the coal mine with more accidents,” he said.
The war over artificial intelligence models has intensified on several fronts, most recently with Chinese companies such as ByteDance introducing advanced technologies. Video generation tools. Last year, Chinese AI company DeepSeek The artificial intelligence industry has been shaken upwhich was primarily led by American companies, by presenting a model that competes with the best artificial intelligence technology in the world. OpenAI He was later charged DeepSeek trains its AI on existing technology in ways similar to those described by Google in its new report.
(Disclosure: Ziff Davis, the parent company of CNET, in 2025 filed a lawsuit against OpenAI, alleging that it infringed Ziff Davis’s copyrights in training and operating its AI systems.)