Accusations that Chinese labs cheat by copying US models are misleading. The practice, known as distillation, is common across the industry (including by Elon Musk's xAI) and academia. Now, with Chinese labs dominating open source, American startups are increasingly building on top of Chinese models.

When a company distills knowledge from a competitor's AI, it's not just scraping pre-training data. It's a highly efficient process of extracting the model's intelligence, reasoning patterns, and skills. This is more akin to an apprentice directly interacting with and learning from a world-class expert than simply reading the same textbooks the expert used.

Leading AI labs, despite intense competition, are collaborating through the Frontier Model Forum to detect and prevent Chinese firms from creating imitation models. This rare alliance is driven by the shared existential threat that 'adversarial distillation' poses to their business models and to U.S. national security.

Proprietary labs argue against 'distillation' (using their model outputs for training) while they have built their own models on vast amounts of copyrighted data. This opposition is an anti-competitive tactic, as model outputs are not copyrightable and distillation helps smaller, open players to compete.

As more of the internet and code repositories are generated by leading AI models, any new model trained on this public data inadvertently "distills" the knowledge and quirks of those proprietary systems. This blurs the line between original training and outright copying.

The battle against AI model distillation is not a niche issue. Anthropic is shutting down millions of accounts per week attempting to distill their models, revealing a highly organized and distributed effort by competitors. This frames the problem as a major cybersecurity and national security challenge, not just a terms-of-service violation.

A new form of analysis compares the semantic similarities (e.g., diction, phrasing) of outputs from different AI models. This technique is being used to create 'fingerprints' that can suggest if one model was illicitly 'distilled' or trained on the outputs of another, a key concern in the AI arms race.

US officials and AI labs allege Chinese firms are engaged in industrial-scale IP theft. They reportedly use fraudulent accounts to extract capabilities from US models like Claude to train their own, creating a facade of domestic innovation.

A key reason for restricting access to new AI models is the threat of 'distillation.' Malicious groups can use thousands of consumer accounts to systematically query a model, effectively reverse-engineering its capabilities. This 'professionalized fraud' can then be used to create powerful open-source alternatives, undermining the entire closed-source business model and security strategy.