Alibaba stock falls 4.9% after Anthropic AI accusation

Alibaba shares dropped as much as 4.9 per cent to a 16-month low in Hong Kong on Thursday after Anthropic accused the Chinese technology group of using fraudulent accounts to draw capabilities from Claude.

The sell-off widened Alibaba’s loss for the year to 33 per cent, according to Bloomberg’s reporting. It also gave investors a market price for a policy problem that had mostly lived in hearings, export-control notices and company statements: who gets access to the most advanced US AI models, and how that access is policed.

Reuters reported that Anthropic said the campaign involved 28.8 million exchanges with Claude and almost 25,000 fraudulent accounts. The company called it the largest known distillation attack it had identified. CNBC reported that Anthropic set out the accusation in a letter and cast the episode as a test of how model makers and governments respond when access controls fail.

The complaint is not a fight over marketing claims or benchmark scores. Anthropic is alleging that its model was queried at scale through false access points, a charge that gives Washington a more concrete reason to revisit how American AI services are monitored abroad.

“We believe combating the threat of illicit distillation requires coordinated action between government and industry, and we will continue working with Congress and the Administration to maintain American AI leadership.”

Anthropic spokesperson, to CNBC

That appeal to Congress and the administration moved the dispute beyond a routine corporate complaint. Traders did not have to wait for a regulator to open a file before adjusting their view of the risk.

Robert Lea, an analyst at Bloomberg Intelligence, said in Bloomberg’s report that Chinese AI models face an elevated risk of a US ban. His read helps explain why one allegation could hit Alibaba’s shares so quickly: it tied a technical dispute to the legal access that underpins a public company’s AI plans.

For the market, the accusation opened several questions at once. US officials could demand tighter screening of overseas accounts. Cloud partners may face more scrutiny. A dispute that begins with model-security controls could still harden into a sanctions or access fight, even if that takes months.

Why investors reacted

The stock move suggested traders were pricing more than the allegation. They were also pricing the chance that model-security claims become an enforcement template for American AI services, cloud access and intermediary accounts. If US officials decide the accusation shows current controls are too loose, companies linked to Chinese AI development could face a harsher compliance setting before any formal ban is announced.

That is a different category of risk.

For Alibaba, the immediate headline is only part of the problem. Even without a formal finding, global investors can mark down the company if they think Chinese technology groups will face tougher routes to chips, cloud capacity and frontier-model access. In that reading, the sell-off reflected regulatory tail risk as much as reputational damage.

The episode landed as investors were already treating AI infrastructure, access rights and export controls as valuation inputs. That made the Hong Kong move larger than a routine controversy over model training practices. It supplied a live example of a question markets have been circling for months: how quickly an AI access dispute can reach the share price of a major Chinese technology company.

For now, the answer is very quickly. Anthropic’s claims have not become a formal enforcement action, but Alibaba’s 16-month low showed that the gap between an AI-security allegation and a market consequence is getting narrower.