OpenAI, Anthropic, and Google collaborate to share intelligence to counter Chinese AI model distillation

According to 1M AI News monitoring, OpenAI, Anthropic, and Google are sharing information through the industry nonprofit Frontier Model Forum, jointly detecting “adversarial distillation” that violates their service terms. The organization was jointly founded in 2023 by the three companies above and Microsoft. Distillation is a technique that trains a new model using the outputs of existing models, replicating the former’s capabilities at lower cost. AI companies themselves also commonly use distillation to create smaller, more efficient versions of their models, but unauthorized third-party distillation has been controversial in the industry.

OpenAI confirmed that it participated in this collaboration and cited a memorandum it previously submitted to the U.S. Congress, saying DeepSeek is trying to “ride on the coattails” to obtain capabilities developed by OpenAI and other U.S. frontier labs. Google, Anthropic, and Frontier Model Forum all declined to comment. U.S. government officials estimate that unauthorized distillation causes losses of billions of dollars in annual profits for Silicon Valley labs.

This collaboration borrows practices from the cybersecurity industry, where companies share attack intelligence among themselves. An AI action plan issued by the Trump administration last year also called for establishing dedicated information-sharing and analysis centers. However, because the boundaries of what can be shared under current antitrust laws remain unclear for each company, the scope of information sharing is limited for now. The three companies have not yet provided public evidence showing to what extent Chinese model innovation depends on distillation, but they say such behavior’s frequency can be measured by the number of large-scale data requests.