In a dramatic escalation of the global AI arms race, Anthropic, the San Francisco-based AI safety company, has accused Chinese tech giant Alibaba of stealing its proprietary model architecture. The accusation, filed with the US International Trade Commission, alleges that Alibaba’s Qwen2.5 model family was trained using trade secrets misappropriated by former employees. This comes as the UK government announces a sweeping reform of intellectual property laws specifically targeting AI-generated works, aiming to create a legal framework that balances innovation with creator rights.
The Anthropic-Alibaba dispute centres on the claim that Alibaba’s models exhibit suspicious structural similarities to Anthropic’s Claude series. Anthropic’s CEO Dario Amodei said the company has evidence of “systematic copying” of model weights and training methodologies. Alibaba has denied the allegations, calling them “groundless” and vowing to defend itself. The outcome could set a precedent for how AI models are protected as trade secrets versus open-source ideals.
Meanwhile, the British government’s new Intellectual Property Bill introduces a three-tier system for AI creations: fully human-authored works retain standard copyright, AI-assisted works get a shortened protection term, and fully autonomous AI outputs are placed in the public domain. This avoids the US-style legal vacuum where AI-generated art has unclear status. The bill also mandates that companies disclose when AI is used in creative processes, a transparency measure that tech lobbyists are fighting.
This dual development underscores a truth I have long argued: the genie of generative AI will not be put back in the bottle, but we must design the bottle’s legal shape with care. Britain’s approach is pragmatic, recognising that total protectionism would stifle innovation, while complete openness risks devaluing human creativity. The Anthropic case serves as a warning that the incentive structures for AI development are fragile. When copying yields profits faster than creation, the entire ecosystem suffers.
For the average user, these battles may seem abstract. But they determine whether your next favourite app is developed by a company that respects the law or one that cuts corners. The user experience of society depends on trust in the digital bedrock. A cloned model might save R&D costs but introduce hidden biases or security flaws. Britain’s moves signal that the rule of law can adapt to the quantum age.
As a technology optimist who has seen the Valley’s hubris first-hand, I welcome both the accusation and the legislation. The former shines a light on the dark arts of corporate espionage; the latter provides a path forward. What concerns me is the pace. While lawyers argue, Chinese labs are cloning, and British artists are suing. We need international norms, not just national laws. Otherwise, the AI future will be a digital Wild West where the outlaws have the fastest algorithms.
In the end, this story is about more than two companies or one country. It is about whether we can shape AI to serve humanity, or whether we will let it divide us. The answer lies not in code but in courts and parliaments. And right now, the verdict is still out.
