In a dramatic escalation of the global AI arms race, Anthropic, the safety-focused artificial intelligence lab founded by former OpenAI researchers, has accused Chinese e-commerce giant Alibaba of stealing its proprietary technology. The accusation, filed in London’s High Court, marks the first major test of British intellectual property laws in the context of advanced AI systems. The case threatens to reshape how algorithmic trade secrets are protected and could set a precedent for the entire industry.
At the heart of the dispute is Anthropic’s constitutional AI framework, a method for aligning large language models with human values. Anthropic alleges that Alibaba’s Qwen-Plus model reproduces core elements of this framework, including specific reinforcement learning techniques and reward model architectures. The company claims it discovered the infringement after running internal deep analysis on Qwen’s publicly available outputs. “We found signatures of our technology in their weights,” said a source close to the investigation. “It is not coincidence. It is cloning.”
Alibaba has strongly denied the allegations, calling them “baseless and an attempt to stifle competition.” In a statement, the company argued that their AI development is independent and built on open-source research. “Anthropic does not own the fundamental maths of machine learning,” the statement read. “Our engineers have never accessed their proprietary systems.”
The case presents novel questions for British courts. Unlike traditional software, AI models are not straightforwardly protected by copyright or patents. Trade secret law, which guards confidential business information, may be the primary vehicle for Anthropic’s claim. However, proving that Alibaba acquired the secrets unlawfully will be difficult. “You cannot simply reverse-engineer a neural network and say it is theft,” explained Dr. Eleanor Shaw, an IP law expert at Cambridge University. “The court will need to decide if the similarities in behavior constitute copying of protected methods, or independent innovation.”
The choice of London as the battleground is strategic. The UK has positioned itself as a global hub for AI regulation, with the government pushing forward the Advanced AI Act that emphasises safety and transparency. A ruling in Anthropic’s favour could encourage other AI labs to litigate here, turning British courts into a forum for tech disputes. Conversely, a dismissal could send a signal that trade secrets are easily bypassed across borders.
Geopolitical tensions underpin the case. The US-China tech decoupling has already seen restrictions on chip exports and cloud computing access. Now, intellectual property claims may become another front. Alibaba’s Qwen models have been widely adopted across Asia, and a finding of infringement could force the company to retrain its models or face a UK sales ban. For Anthropic, which has focused on ethical AI since its founding, the lawsuit is also about damaging a competitor perceived as less safety-conscious.
Broader implications for the AI ecosystem are profound. Many AI models are built on research published in open papers, making it hard to draw clean lines between inspiration and theft. “We are moving from a world of academic collaboration to one of corporate litigation,” noted Julian Vane, Technology & Innovation Lead. “Every lab will now have to patent every incremental improvement or risk having it lifted. This could slow progress and entrench incumbents.”
The hearing is expected to last six months, with technical evidence from both sides dissecting model architectures and training data. For now, the AI community watches with bated breath. The verdict may not just decide the fate of two companies, but shape the future of innovation itself.
