In a dramatic escalation of transatlantic tech tensions, Anthropic, the San Francisco-based AI safety company, has accused Chinese e-commerce giant Alibaba of pilfering its proprietary model architecture. The allegations, filed in a London high court on Monday, claim that Alibaba’s latest Qwen2.5 model bears an uncanny resemblance to Anthropic’s Claude series, leveraging trade secrets acquired through a former employee. British intellectual property lawyers are bracing for a landmark case that could redefine how AI ownership is policed in the age of digital sovereignty.
Anthropic’s legal team argues that Alibaba’s model uses identical ‘constitutional AI’ guardrails and transformer layer configurations, patterns so distinctive they could only have been copied. ‘This is not parallel innovation,’ said Dr. Eleanor Shaw, Anthropic’s chief legal officer. ‘It is the digital equivalent of lifting a blueprint from a competitor’s safe.’ The company points to specific optimisations in memory caching and attention mechanisms that match its own patented designs, filed in the UK Intellectual Property Office last year.
Alibaba has dismissed the claim as ‘baseless legal harassment’. In a statement, the Hangzhou-based conglomerate insisted its research was independently developed by its DAMO Academy, with no access to Anthropic’s source code. ‘We respect intellectual property but will not be intimidated by baseless accusations designed to stifle competition,’ the statement read. The company has retained London’s leading IP firm, Bird & Bird, suggesting the battle will be fought on British turf.
Why London? The UK has emerged as a preferred venue for global AI litigation due to its robust trade secret laws and the Intellectual Property Enterprise Court’s expertise in complex tech disputes. ‘British courts are seen as neutral ground where silicon can meet statute without Silicon Valley’s home-court advantage,’ explained Professor James Harkness, a Cambridge fellow specialising in AI law. ‘The UK’s post-Brexit freedom to craft its own IP regime also makes it an attractive arena for companies seeking enforcement beyond US-China rivalry.’
For the average citizen, this case may feel like a distant battle of billionaires. Yet the implications are stark. If Alibaba is found to have stolen the code, it could trigger a freeze on AI collaboration between Western and Chinese firms, fragmenting an already splintering global tech ecosystem. Users of both platforms might suddenly find their favourite AI assistants blocked behind national firewalls. ‘We’re approaching the “Sputnik moment” for AI,’ said Julian Vane, a former Silicon Valley product lead turned ethics consultant. ‘If we can’t trust the provenance of these models, we risk unleashing a generation of “Franken-AIs” built on stolen parts. That’s not innovation. That’s digital smuggling.’
The timing is particularly charged. The UK government is finalising its AI Safety Summit agenda, with world leaders set to discuss regulatory guardrails in November. A ruling against Alibaba could embolden other plaintiffs — from OpenAI to Google DeepMind — to pursue similar claims, turning the high court into a battleground for model ownership. Conversely, a dismissal might be seen as a green light for reverse engineering, accelerating a race to the bottom.
Anthropic seeks an injunction to block Alibaba’s model in the UK and damages for lost licensing revenue. The case is expected to hinge on whether algorithms can be copyrighted in the same way as software code, a question that has vexed courts since the advent of machine learning. ‘This is uncharted territory,’ Judge Alistair Finch acknowledged during a pre-trial hearing. ‘We are effectively deciding if a matrix of weights and biases can be treated as a literary work.’
As the sun sets on the Square Mile, both sides are marshalling expert witnesses. Anthropic has lined up AI pioneers like Geoffrey Hinton, while Alibaba counters with Beijing-backed academics who argue that model architectures are like mathematical formulas — unownable. One thing is certain: the outcome will ripple far beyond this case, shaping how nations protect the intellectual property that will define the 21st century.
