OpenAI, Anthropic & Google Unite Against Chinese AI Model Theft

For the first time in the industry’s history, three rival AI giants — OpenAI, Anthropic, and Google — have joined forces to combat a growing threat: the unauthorized copying of their proprietary AI models by Chinese tech companies. The coordinated effort, announced through the Frontier Model Forum, marks a new phase of geopolitical tension in the global AI race.

What Is Adversarial Distillation?

At the heart of the controversy is a technique known as adversarial distillation. Rather than building a model from scratch, an attacker generates massive volumes of outputs from a target model — like OpenAI’s GPT series or Anthropic’s Claude — and uses those outputs to train a copycat model at a fraction of the cost. The process is technically against the terms of service of every major AI provider, but it’s difficult to detect without large-scale monitoring.

The critical danger, beyond intellectual property theft, is that distilled models don’t inherit the safety work baked into the originals. Alignment techniques, content filters, and refusal training don’t transfer cleanly through distillation — leaving copycat models potentially more dangerous and less controllable than their sources.

What Happened

According to details shared by Anthropic, three Chinese AI companies — DeepSeek, Moonshot AI, and MiniMax — collectively generated over 16 million fraudulent exchanges with Claude through approximately 24,000 illegally created accounts. The scale of the operation suggests a coordinated, systematic effort to extract training data rather than isolated incidents of misuse.

In response, OpenAI, Anthropic, and Google announced on April 6–7, 2026 that they are collaborating through the Frontier Model Forum — an industry nonprofit the three companies co-founded in 2023 alongside Microsoft — to share threat intelligence and detect adversarial distillation attempts in real time. This is the first time the Forum has been activated as an active threat-intelligence operation against a specific external adversary.

The move is notable not just for its scale, but for who’s involved. OpenAI, Anthropic, and Google are fierce competitors racing to lead the next generation of AI. That they’ve set aside that rivalry to coordinate against a common threat signals just how seriously the industry views model IP theft. Reporting on the alliance has appeared across Bloomberg and other major outlets.

Why It Matters

The stakes go beyond corporate competition. Distilled models built from frontier AI without safety alignment layers represent a genuine risk — models capable of advanced reasoning but without the guardrails that took years and hundreds of millions of dollars to develop. If adversarial distillation becomes a viable shortcut for state-backed or commercial actors, the entire AI safety ecosystem faces a structural challenge.

On the geopolitical front, this is the latest flashpoint in the US-China AI competition. Regulators in Washington and Brussels will be watching closely — the alliance’s next move, whether that includes legal action, trade policy advocacy, or new technical countermeasures, could shape how international AI development is governed for years to come.

What to Watch Next

Keep an eye on whether the Frontier Model Forum pursues legal action against the named companies, and how DeepSeek, Moonshot AI, and MiniMax respond publicly. It’s also worth monitoring whether other major model providers — Meta, Mistral, or xAI — join the coalition. As the AI industry matures, intellectual property enforcement may become as critical as the models themselves.

For the full background on the alliance’s formation, see the original reporting at Bloomberg and BanklessTimes. Continue reading for the latest developments in AI security and geopolitics.