OpenAI, Anthropic and Google have begun working together to clamp down on Chinese competitors extracting and distilling their AI models. The three companies — fierce rivals in the domestic market for AI talent, users and enterprise contracts — are sharing information about unauthorized model copying in what Bloomberg described as "an unusual area of cooperation."[2]

The alliance is not formal. There is no treaty, no joint venture, no shared headquarters. It is an intelligence-sharing arrangement: when one company detects a Chinese firm distilling its model, it shares the evidence with the others. The goal is to build a common defense against what the companies describe as both an economic threat and a national security concern.[1]

The Distillation Threat

Model distillation is the process of training a smaller AI model on the outputs of a larger one. A Chinese company can query OpenAI's GPT-5, record the responses, and use those responses to train a competing model that costs a fraction of the original to develop. It is the AI equivalent of reverse engineering — legal in some contexts, illegal in others, and difficult to prove in all of them.[3]

The three companies believe that Chinese firms are using distillation at scale, creating competing models that undercut Western pricing while benefiting from Western research investment. The economic losses are difficult to quantify, but the competitive threat is clear: if a Chinese company can produce a model that is 80 percent as good at 20 percent of the cost, the Western companies' pricing power evaporates.[4]

The Irony

The same three companies are simultaneously laying off tens of thousands of workers. Oracle cut 30,000. Amazon cut 30,000. Intel cut 25,000. Microsoft cut 15,000. The total for Q1 2026 exceeds 90,000 tech jobs — roughly 973 per day. The companies are gutting their workforces while investing billions in AI infrastructure, and now they are uniting against a foreign competitor that they accuse of stealing the very technology they are using to justify the layoffs.

The irony is not lost on the workers who lost their jobs to AI. They were told that automation was inevitable. Now the companies that automated them are asking the government to protect them from foreign competitors who are doing the same thing.[6]

The National Security Angle

The companies cite national security as a motivation for cooperation. If Chinese firms can distill Western AI models, they can also distill models used for intelligence analysis, cyber defense and military planning. The concern is not just economic — it is strategic.[5]

The US government is aware of the cooperation. Whether it is directing it is unclear. But the alignment between corporate and national security interests is unmistakable: the companies that build America's AI infrastructure are the same companies that are asking the government to protect that infrastructure from foreign competition.

The Precedent

This is the first time that OpenAI, Anthropic and Google have cooperated on any significant scale. They compete for talent, for enterprise contracts, for research prestige and for the attention of regulators. Their cooperation on model copying suggests that the threat from China is significant enough to override domestic rivalry.

It also suggests that the AI cold war is no longer a metaphor. It is a reality — and the first battle is over who gets to own the models that will define the next decade of computing.

-- DAVID CHEN, Beijing