Anthropic has accused Chinese technology giant Alibaba of carrying out what it describes as the largest known attempt to extract the capabilities of its Claude AI model, escalating concerns over how leading AI systems are being copied and replicated. In a letter sent to US lawmakers, the company alleged that operators linked to Alibaba conducted nearly 29 million interactions with Claude using thousands of fraudulent accounts in an effort to harvest the model's knowledge and behaviour.Key TakeawaysAnthropic has accused Alibaba-linked operators of conducting a large-scale AI model extraction campaign.The company claims nearly 29 million exchanges were made with Claude using fraudulent accounts.Anthropic says the activity involved "distillation attacks" designed to replicate Claude's capabilities.The AI firm has urged US lawmakers to strengthen protections against model theft.Alibaba has not publicly responded to the allegations and has previously denied links to China's military.Anthropic says Claude was targeted through large-scale extraction effortsThe dispute boils down to what Anthropic calls a targeted attempt to steal key features from its Claude AI model. In a letter sent to Senators Tim Scott and Elizabeth Warren on June 10, Anthropic claims that people linked to Alibaba set up thousands of fake accounts and queried Claude almost 29 million times.Anthropic says these efforts focused on breaking down and copying some of Claude’s most sophisticated skills - things like complex reasoning, understanding long stretches of context, and decision making. They describe this as the biggest data extraction campaign they’ve ever seen and warn that it’s part of a larger problem facing AI companies pouring resources into next-generation models.All this comes as AI companies around the world are locked in fierce competition, racing to build better models while trying to keep costs under control.Distillation attacks have become a growing concern for AI companiesAt the centre of Anthropic's claims are so-called distillation attacks. The practice involves using outputs generated by a powerful AI model to train another model that may not possess the same capabilities independently.According to Anthropic, these attacks allow companies to shortcut years of research and development by effectively learning from a more advanced system. The company argued that such activity can transform massive investments in AI development into an unintended advantage for competitors.Anthropic further claimed that these extraction efforts are being conducted on an industrial scale and are designed to enable rival companies to recreate sophisticated AI capabilities at a fraction of the cost required to develop them from scratch.The issue is not unique to Anthropic. Other major US AI developers have previously raised concerns about similar techniques being used to accelerate AI development outside the United States.The dispute adds another layer to US-China technology tensionsThe allegations arrive against the backdrop of growing technology competition between Washington and Beijing, particularly in areas such as artificial intelligence and advanced semiconductors.In its letter, dated 10th June, Anthropic argued that model extraction campaigns have implications beyond commercial competition and could affect national security interests. The company referenced claims from the US Department of Defence that several Chinese firms, including Alibaba, have ties to the Chinese military.Alibaba and other companies have previously denied such allegations. The report also notes that Alibaba recently filed legal action seeking removal from a Pentagon blacklist.Anthropic has urged Congress to introduce stronger penalties for companies involved in model extraction efforts and to implement additional safeguards designed to prevent advanced American AI technologies from being replicated by competitors.AI developers face a new challenge as models become more valuableThe fight underscores a bigger quandary for AI firms as their models grow more potent and commercially valuable. Unlike traditional software, AI systems can potentially leak large amounts of information through repeated interactions, creating opportunities for competitors to study and mimic their behaviour.For companies like Anthropic, safeguarding the model’s abilities is becoming nearly as important as developing them. The company says that without stronger safeguards, economic incentives to build advanced AI systems could be undermined.The controversy comes at a critical time for Anthropic, which is considered one of the top AI creators alongside OpenAI.As competition intensifies and AI models become more powerful, questions around intellectual property, model security and the limits of AI training practices are likely to remain at the centre of industry debates.Frequently Asked Questions1. How to Detect Model Extraction Attacks from AI Companies?AI developers monitor for anomalous usage patterns, large-scale automated interactions, suspicious account activity and repeated querying behaviour that could indicate attempts to replicate model capabilities.2. Why is AI model distillation a controversial topic?The method could dramatically cut the time and cost of building advanced AI systems, leading developers to claim it can undercut years of research and investment.3. Would claims like these result in tighter AI regulation in the US?Maybe. Maybe. Anthropic had previously urged lawmakers to consider stronger penalties and safeguards to prevent the unauthorised extraction of advanced AI capabilities.4. Why are frontier AI models seen as valuable intellectual property?Modern AI models require huge investments in computing infrastructure, data processing, and research, rendering their capabilities among the most valuable assets of AI companies.5. Is this the first time that a US AI company has raised concerns about model copying?No. Several AI developers, including OpenAI, have raised concerns in the past about rivals using the outputs of advanced models to train competing systems.end of article