Anthropic, the San Francisco‑based AI research firm behind the Claude series of language models, sent a sharply worded letter to Congress this week demanding that the United States take decisive action against Alibaba Group. The letter accuses the Chinese e‑commerce giant of "brazenly" racing to produce a copycat version of Claude, a move the company says directly contravenes a recent Trump administration memorandum that labeled foreign model‑cloning attempts "unacceptable."
According to Anthropic, Alibaba’s activities unfolded just weeks after the memo was issued, despite the fact that the Chinese conglomerate is listed on the New York Stock Exchange, operates a substantial business presence in the United States, and answers to U.S. investors and regulators. The firm argues that Alibaba’s conduct not only flouts U.S. policy but also accelerates China’s ability to achieve capabilities comparable to Anthropic’s upcoming Mythos Preview release.
Anthropic’s letter does not stop at condemnation; it lays out a three‑part legislative agenda aimed at curbing China’s progress in AI. First, the startup urges Congress to modernize antitrust law so that American AI companies can share intelligence about evolving Chinese tactics without fear of legal repercussions. Second, it calls for expanded export controls on high‑performance chips, arguing that limiting China’s access to advanced compute would render distillation attacks—where a model is trained on the outputs of another—far less effective. Finally, Anthropic recommends penalties that would make it more costly for Chinese labs to rely on U.S. models or foreign data centers, including restrictions on accessing U.S. models, advanced chips, or overseas cloud services.
Alibaba, for its part, has not commented on Anthropic’s accusations. The company is simultaneously locked in a legal battle with the Trump administration. In a lawsuit filed Tuesday, Alibaba alleges that the administration unfairly blacklisted the firm by falsely linking it to the Chinese military. The lawsuit seeks to have the designation removed, contending that Alibaba’s board is independent, its products focus on retail, logistics, and enterprise IT, and it has no involvement in weapons, defense, or intelligence work.
While Anthropic’s demands are clear, the path to legislative action remains uncertain. The Trump administration’s memo, issued in early 2024, warned that cloning U.S. frontier models was “unacceptable,” but it stopped short of detailing enforcement mechanisms. Anthropic’s push for stronger penalties reflects growing anxiety in the U.S. AI community about a perceived “AI arms race” with China, where model‑stealing and distillation could erode the competitive edge of American firms.
Industry observers note that the clash highlights a broader tension: U.S. AI firms rely heavily on Chinese manufacturing and semiconductor supply chains, yet face mounting pressure to protect intellectual property from foreign replication. Anthropic’s appeal to Congress underscores the dilemma of balancing open innovation with national security concerns.
As the debate unfolds, policymakers will have to weigh the potential impact of stricter antitrust rules and export controls against the risk of stifling U.S. AI development. For now, Anthropic’s letter adds a new voice to the chorus urging Washington to act before China narrows the gap in frontier AI capabilities.
Questo articolo è stato scritto con l'assistenza dell'IA.
News Factory APP - notizie agentiche per potenziare il tuo SEO e AEO.