Anthropic report alleges distillation campaigns by Alibaba, Moonshot AI, and DeepSeek

Anthropic released a report on Thursday alleging that Alibaba, Moonshot AI, and DeepSeek have run persistent distillation campaigns against its models, and that the activity has escalated in recent months as competition among AI developers has intensified. TechCrunch reported the release, and its account identifies all three companies as China-based AI companies.

Distillation in this setting means training a model on the outputs of another model, which lets a developer approximate a stronger system's behavior without access to its weights. The report Anthropic released characterizes the campaigns it describes as attacks, and TechCrunch's account presents them as persistent rather than one-off incidents. The excerpt does not explain how Anthropic detects or attributes such activity.

The specifics behind the allegations are not in the material available. TechCrunch's account gives no count of accounts, prompts, or model exchanges, names no Anthropic model or service as the target, and sets no start date for the alleged campaigns. The one time frame it provides is recent months, tied to the company's claim that the activity has escalated. A figure for the scale of the campaigns would have to come from the report itself, which is not quoted at length in the excerpt.

Anthropic stands as the source of the allegations as presented. The account does not say which systems detected the activity, how the company tied it to individual labs, or whether it notified the named companies before publishing. Alibaba, Moonshot AI, and DeepSeek are identified only as China-based AI companies; none is quoted, and no response from any of them appears in the material available. No independent body is named as having verified the claims, and the story's use of alleges reflects that they remain Anthropic's assertions.

Anthropic links the alleged escalation to market pressure. It attributes the increase to intensifying competition in the sector, which frames the campaigns as part of a broader race among AI developers rather than an isolated case. That causal claim is Anthropic's characterization, and the excerpt supplies no measurement behind it, no incident log, and no explanation of how the company separates targeted extraction from ordinary API traffic.

The form of the report is unstated in the account. TechCrunch does not say whether Anthropic published it as a blog post, a technical paper, or a filing, whether supporting data accompanied it, or whether the company shared it with regulators or customers. Those details would determine how much of the underlying evidence is available for outside review.

Only the TechCrunch report is available here, so there is no second account to corroborate or dispute Anthropic's claims and no reaction from a regulator, competitor, or customer. Anthropic's earlier statements on distillation are not part of this source set, which leaves open how the new allegations relate to anything the company said before.

The account sets out no next step. It names no enforcement action, no deadline for the named companies to respond, and no signal that Anthropic has referred the matter to authorities. The open questions are the evidence behind the attribution, the volume of activity Anthropic says it measured, and whether Alibaba, Moonshot AI, or DeepSeek will contest the account.