Anthropic claims Alibaba defied Trump to attack Claude and steal capabilities (arstechnica.com)

Anthropic has accused Alibaba of orchestrating a significant attack to clone its AI model, Claude. This operation, described as the largest of its kind, reportedly involved more than 28.8 million exchanges through nearly 25,000 fraudulent accounts, targeting Claude's advanced capabilities like agentic reasoning and software engineering. The allegations emerged in a letter to U.S. Senators ahead of a hearing on AI, highlighting concerns over China's attempts to catch up with U.S. AI advancements without investing in extensive research and development. Anthropic emphasizes that this campaign not only violates their terms of service but also reflects a growing trend where Chinese companies exploit American innovations. The implications of this incident are substantial for the AI/ML community, as it sheds light on the increasing risks of IP theft and model cloning in the race for AI supremacy. Anthropic's warnings about a burgeoning "circumvention economy" signal potential future threats to the integrity of AI development and investment in the United States. This scenario underscores the ongoing geopolitical tensions in technology and raises critical discussions on how to safeguard innovation against illicit practices, particularly in light of previous warnings from the U.S. government about industrial-scale AI theft from China.