OpenAI, Anthropic, and Google have done something they almost never do: agree. The three companies, which compete viciously for the same customers, talent, and compute resources, are now sharing intelligence through the Frontier Model Forum to detect and counter what the industry calls adversarial distillation. Translation: Chinese companies are using their AI models to build cheaper knockoffs, and it is costing Silicon Valley billions.

Bloomberg first reported the collaboration, which is channeled through the nonprofit the three companies co-founded with Microsoft back in 2023. US officials have estimated that unauthorized distillation costs American AI labs billions of dollars in lost annual profit. The technique itself is straightforward: a "student" model extracts knowledge from a "teacher" model at a fraction of the original development cost. AI labs use it themselves all the time. The problem is when adversary nations do it without permission.

The alarm bells started ringing in January 2025 when DeepSeek released its R1 reasoning model. The system was shockingly capable, and it immediately raised questions about how it was built. Microsoft and OpenAI investigated whether DeepSeek had improperly extracted large volumes of data from OpenAI's models. OpenAI subsequently told US lawmakers that DeepSeek had used "increasingly sophisticated tactics" to pull results from its models, accusing the Chinese firm of free-riding on American research.

Each company has responded differently. Anthropic has built what it calls a "trust and safety" pipeline to detect distillation patterns in API traffic. Google has tightened its terms of service and is using fingerprinting techniques to identify when Gemini outputs are being used to train competing models. OpenAI has been the most aggressive, going directly to Congress with evidence.

Here is the awkward part. These three companies cannot agree on pricing, safety standards, open source policy, or whether AI should be used by militaries. They have spent the last two years trying to destroy each other commercially. Anthropic positions itself as the safety-first alternative to OpenAI. Google uses open source to undercut both. OpenAI lobbies against everyone. But China? China unites them instantly.

The geopolitical framing is important. This is not just about intellectual property theft. It is about who controls the foundational technology of the next century. If Chinese labs can replicate frontier model capabilities at a fraction of the cost by extracting from American models, the entire economic moat that justifies $852 billion valuations starts to look like a sand castle.

My read: this alliance will hold exactly as long as the China threat feels existential. The moment it recedes, these companies will go back to fighting each other. But for now, the enemy of my enemy is my business partner. And that might be the most honest thing Silicon Valley has done in years.