OpenAI, Anthropic, and Google have reportedly formed a joint effort to counter adversarial model distillation, with particular attention on activities linked to Chinese AI companies. According to Bloomberg, the three firms will share intelligence through a “cutting-edge models forum” designed to identify and respond to suspected distillation attempts targeting leading AI systems.
The reported alliance follows earlier claims from Anthropic that some companies created fraudulent accounts to extract outputs from its models and use that data to train competing systems. The report specifically named DeepSeek and MiniMax, adding to broader tensions in the AI sector over intellectual property, platform misuse, and competitive boundaries.
Why model distillation has become controversial
Model distillation generally refers to techniques used to reproduce some of the capabilities of advanced systems at a lower cost. In many settings, distillation is a legitimate technical method for improving efficiency and reducing deployment expenses. However, when a company relies on another provider’s commercial service to harvest outputs and build a rival model, that behavior is often classified as adversarial distillation and is typically banned under platform terms.
For companies such as OpenAI and Anthropic, the issue is not only about protecting expensive research and development. They also argue that improperly distilled models could inherit useful capabilities while bypassing the original guardrails, safety restrictions, or monitoring systems built into the source platforms.
What the alliance is trying to achieve
By coordinating more closely, the three US AI firms appear to be trying to protect their frontier-model investments, improve detection of suspicious usage patterns, and reduce the risk of low-cost replication by competitors. As the generative AI race becomes more intense and model-building costs remain high, disputes over data extraction, imitation, and security controls are becoming central industry issues.
So far, the report does not provide further operational details about how the partnership will work in practice, and no additional response from the companies named in the allegations was included. Even so, the legality and limits of model distillation are likely to remain a major point of debate across the global AI industry.