OpenAI, Google and Anthropic exchange information to detect China’s attempts to copy their AI models

OpenAI, Google y Anthropic have joined forces to combat attempts to distill artificial intelligence models by Chinese companies, given the appearance of imitations of their AI products.

The three companies are sharing information through the Frontier Model Forum, an initiative founded by themselves in 2023 to ensure the safe and responsible development of frontier AI models.

The intention now is detect adversarial distillation attemptsespecially those from Chinese companies, as sources familiar with this matter have indicated to Bloomberg. It is a technique that seeks to clone or replicate the functionality of a model in a smaller one to obtain its capabilities but at a lower cost.

This technique is a violation of the terms of use of the models of the three companies, and also has economic repercussionssince the unauthorized use of distillation is estimated at billions of dollars annually, according to one of the sources consulted.

OpenAI has confirmed this joint work to Bloomberg. The company already issued a letter to the United States Congress in February in which it directly accused the Chinese company DeepSeek to clone American models, including their own.

Some components of the Goal Flame models, distributed under an open source license, have been found in the DeepSeek R1 reasoning model, which made the company rethink its strategy with the launch of new models.