A recent letter sent to the U.S. Senate Committee on Banking, Housing, and Urban Affairs has shed light on a significant AI distillation attack allegedly carried out by Chinese tech giant Alibaba against Anthropic, a leading AI company. According to the letter, Alibaba's AI lab utilized approximately 25,000 fraudulent accounts to engage in 28.8 million exchanges with Anthropic's models between April 22 and June 5, in what Anthropic describes as the largest known distillation attack on their systems to date. This technique, known as distillation, involves training a smaller model using the outputs of a more powerful existing model. Anthropic's spokesperson emphasized the need for coordinated action between government and industry to combat the threat of illicit distillation, underscoring the importance of maintaining American AI leadership. The incident comes on the heels of a White House Office of Science and Technology Policy memorandum aimed at helping AI companies detect and counter industrial-scale distillation attacks. Notably, this is not Anthropic's first encounter with such attacks, as the company previously identified three other industrial-scale distillation campaigns from different AI labs. The situation is further complicated by a recent export control directive from the Trump administration, which has ordered Anthropic to suspend access to its latest models for foreign nationals, citing national security concerns. As the situation unfolds, Anthropic is working closely with policymakers to resolve the issue and ensure the continued development of its AI capabilities.

Source: CNBC