Anthropic accuses Alibaba of world's biggest AI theft
What's the story
US-based artificial intelligence (AI) company Anthropic has accused Chinese tech giant Alibaba of a massive intellectual property theft. The company alleged that the Chinese behemoth attempted to illegally extract capabilities from its Claude AI models in what is being termed as the biggest known case of such an attempt. The allegations highlight growing concerns over IP theft amid intensifying tech rivalry between US and China.
Allegations
Campaign involved 28 million interactions with Claude
Anthropic has alleged that operators associated with Alibaba and its AI research unit, Qwen, ran a major "distillation" campaign from April 22 to June 5. The campaign allegedly involved over 28.8 million interactions with Anthropic's Claude platform via nearly 25,000 fake accounts. Anthropic said this is the largest-known campaign of AI distillation, aimed at replicating its technology.
Methodology
Understanding 'distillation' process
Distillation is a method where a smaller or less powerful AI model is trained on the outputs of a more powerful one. While this process is mostly used internally by companies to improve efficiency, it has also been misused by some to replicate proprietary capabilities without permission. Anthropic accused Alibaba and other Chinese labs of using this technique, called adversarial distillation, to build rival chatbots at lower costs.
Strategic move
Aiming to match Mythos preview model capabilities
Anthropic claimed that the alleged distillation campaign was aimed at speeding up China's efforts to match the capabilities of its advanced Mythos Preview model, one of Anthropic's latest frontier AI systems. The company detailed these allegations in a June 10 letter sent to US Senators Tim Scott and Elizabeth Warren ahead of an artificial intelligence and national security hearing.
Extraction details
China illicitly extracting US AI capabilities, says Anthropic
In the letter, Anthropic said the alleged campaign was conducted "illicitly, systematically, and at an industrial scale" to extract US AI capabilities across frontier labs. The company also claimed these capabilities were repackaged as Chinese-owned products without bearing the training and research costs required to build frontier models. These allegations come after White House Office of Science and Technology Policy director Michael Kratsios warned about such activities in April.
Past incidents
Similar claims made earlier this year
The latest allegations come after similar claims made by Anthropic earlier this year. In a February blog post, the company had claimed that Chinese AI start-up DeepSeek and other Chinese AI developers were trying to extract capabilities from Claude. At the time, Anthropic had warned that these campaigns were becoming more sophisticated and frequent, requiring coordinated action from governments, tech companies, and the wider AI community.
Legislative action
US lawmakers move to address concerns
In response to the growing concerns, US lawmakers are moving to address the industry's worries. Tennessee Republican Bill Hagerty and New Jersey Democrat Andy Kim plan to introduce an amendment to defense legislation that would blacklist or sanction any Chinese firm found improperly accessing US AI model output for training competing models. The move highlights the increasing intersection of AI development with national security issues.