Legal and regulatory

Anthropic report accuses Moonshot of covertly routing Kimi user requests through Claude

2 sources5 of 6 claims verifiedNotable signal
Place in VP50
#20
Sources
2
Claims verified
5 of 6

What happened

AI summaryWritten from 2 sources, checked claim by claim

Anthropic accused Moonshot AI, the maker of Kimi, of secretly routing user requests through its Claude models, in a report covered by Bloomberg and TechCrunch on September 10, 2026. Bloomberg reported that Moonshot diverted user queries to Anthropic's chatbots without telling users. Anthropic said it spotted nearly 300,000 requests, sent primarily to its Opus model. TechCrunch said the requests ran over a 10-day period through a network of 5,000 accounts.

TechCrunch said the report also described distillation campaigns by Alibaba and DeepSeek, with nearly 200 million exchanges observed across five campaigns in total. It quoted Anthropic as saying unauthorized labs had developed increasingly sophisticated methods to circumvent its defenses, a statement that covered all the campaigns. TechCrunch wrote that the Moonshot campaign seemed to route requests directly from the Chinese military, without presenting this as a confirmed finding. These were allegations by Anthropic, not findings by a court or regulator. TechCrunch reported no response from Moonshot and no legal action.

Key points

  1. Anthropic accused Moonshot of routing Kimi user queries to Claude without telling users, Bloomberg reported.1
  2. Anthropic said it spotted nearly 300,000 requests sent primarily to its Opus model.12
  3. TechCrunch said the requests ran over a 10-day period through a network of 5,000 accounts.2
  4. TechCrunch reported no Moonshot response and no legal action.2

Why it matters

Named allegation

Anthropic publicly named Moonshot in a report on distillation campaigns. The claim comes from Anthropic alone and is not a finding by a court or regulator.

Product integrity question

The report, as Bloomberg described it, alleged that some Kimi user requests were answered by another provider's model. That bears on what customers were buying from Moonshot, if the allegation holds.

Legal exposure unresolved

No legal action appeared in the coverage read, so any legal or regulatory cost remains undefined. Moonshot's side was not reported.

In context

$100M$1B$10B202420252026$35B
Moonshot AI priced rounds
RoundDatePost-money valuationRaisedLed by
Angel roundJun 13, 2023$300M$200MUnknown
Series A+Feb 20, 2024$2.5B$1.0BAlibaba
Series BAug 5, 2024$3.3B$300MUnknown
Series CDec 2025$4.3B$500MIDG Capital
Funding roundFeb 2026$10B$700M九安医疗 (Jiuan Medical) and 五源资本 (5Y Capital)
Series FJul 29, 2026$35B$3.5BNational AI Industry Investment Fund
Post-money valuation at each priced round. Log scale.Hover a point for the round

Updated

Fact-check

  • Anthropic's report, covered on September 10, 2026, alleged Moonshot secretly routed Kimi user requests to Claude without telling users.The 'secretly routed user requests through Claude' framing is in [1] Bloomberg's headline in the event data; the Bloomberg page returned 403 to the checker. [2] TechCrunch, dated September 10, 2026, confirms the report and that Moonshot requests were routed to Claude.Partly verified
  • Anthropic spotted nearly 300,000 requests sent primarily to its Opus model.Confirmed by [2] TechCrunch: nearly 300,000 requests routed to Claude, mainly targeting Opus.Verified
  • The requests ran over a 10-day period through a network of 5,000 accounts.Confirmed by [2] TechCrunch.Verified
  • The report covered nearly 200 million exchanges across five campaigns, including Alibaba and DeepSeek, and Anthropic said unauthorized labs had developed increasingly sophisticated methods to circumvent its defenses.Confirmed by [2] TechCrunch, which names Alibaba and DeepSeek, gives the 200 million and five-campaign figures and carries the quote.Verified
  • TechCrunch wrote that the Moonshot campaign seemed to route requests directly from the Chinese military.Confirmed by [2] TechCrunch, which uses the word 'seemed'.Verified
  • TechCrunch reported no Moonshot response and no legal action.Confirmed by the checker's read of [2] TechCrunch, which mentions neither.Verified

Sources

  1. [1]Bloomberg (opens in a new tab)Moonshot Secretly Routed User Requests Through Claude, Anthropic SaysAnthropic said it spotted nearly 300,000 customer requests that Moonshot sent primarily to Anthropic's Opus AI model.
  2. [2]TechCrunch (opens in a new tab)Anthropic details distillation campaigns from Alibaba, Moonshot AI, and DeepSeekunauthorized labs have developed increasingly sophisticated methods to circumvent our defenses

More signals

All signals →