2026-09-30 — OpenAI attributes a reasoning-extraction campaign to Moonshot-linked users

• OpenAI said it disrupted a coordinated campaign to extract its models’ protected reasoning and attributed a core cluster of it to individuals associated with Moonshot AI, maker of Kimi The activity began 1 July, peaked on 24 and 25 July with 16,000 attempted requests from more than 4,000 users, and a cluster of more than 15,000 users was fully disrupted by 28 July; OpenAI says it cannot tell whether all operators were one actor. (OpenAI, Unite.AI)

2026-09-23 — Chinese AI shares fall on the reported probe

• Shares of Chinese AI companies fell after the report that China’s internet regulator is investigating DeepSeek and Moonshot AI, with Alibaba down about 4% Bloomberg’s headline tied the selloff to the report, which undercut the lift from Alibaba’s new-chip announcement. (Bloomberg)

2026-09-22 — China’s regulator reportedly probes DeepSeek and Moonshot over data sent to Claude

• China’s internet regulator is investigating DeepSeek and Moonshot AI over user data that may have reached Anthropic’s Claude, after summoning all seven Chinese companies named in Anthropic’s 09-10 threat report, The Information reported The Cyberspace Administration reportedly questioned staff at both firms, narrowing in on them because of the report’s examples (Moonshot allegedly passed a user’s police-camera footage request to Claude without the user knowing); the probe is open, no penalty is decided, neither company has commented, and The Next Web could not independently verify it. (The Information, The Next Web, Gizmodo)
• China’s Cyberspace Administration is investigating DeepSeek and Moonshot AI over whether user data was sent to Anthropic’s Claude, The Information reported, after summoning executives from the Chinese companies named in Anthropic’s 09-10 threat report and then narrowing its scrutiny to those two. The trigger is Anthropic’s account that Moonshot routed more than 23 million exchanges to Claude between May and July and DeepSeek 12.1 million in 14 days in July, in some cases carrying personal or state-linked data; Beijing’s concern is what flowed out to a US company, the reverse of Anthropic’s grievance about what was taken. The probe rests on The Information’s sources: neither the regulator nor either company has confirmed it, and no penalty is reported. It surfaced two days before the Trump-Xi summit. (Gizmodo, Startup Fortune)
• Xiaomi released MiMo-V2.6-Pro under an MIT licence on 09-22, and Artificial Analysis scores it 46 on its Intelligence Index, the highest of any open-weight model, at about $0.13 per task. Xiaomi credits a reinforcement-learning run of under six days costing about $2.62 million and is releasing its training toolkit; the launch comes 12 days after Anthropic’s threat report named Xiaomi in case GTG-16008, alleging more than 400,000 exchanges routed through Claude over 20 days in March and April to gather training data. (The Next Web, The Decoder)

2026-09-12 — Moonshot denies and files a police report over viral rumors that its founder was detained, in the direct wake of Anthropic’s distillation report; a day earlier, its own revenue target came out untouched by the fallout

• Moonshot AI issued a statement on 2026-09-12 calling online rumors that founder Yang Zhilin and 15 senior executives had been “taken away for investigation” — which spread across Chinese social platforms immediately after Anthropic’s threat report named Moonshot on 09-10/09-11 (below) — “completely fabricated and malicious slander,” saying it had filed a police report and would pursue legal action against whoever is spreading them. No news outlet or government agency has independently confirmed any detention or investigation exists; the rumor and the denial are, so far, the only confirmed facts. This is Moonshot’s first public response of any kind to the distillation fallout, though it answers a rumor about consequences rather than the underlying distillation allegation itself. (BeInCrypto, BigGo Finance)
• A day earlier (2026-09-11, after this map’s prior close), Moonshot told Bloomberg it is targeting roughly $2 billion in annualized revenue by year-end — doubling its August run rate — while Kimi K3 continues generating around 300 billion tokens/day on OpenRouter, usage having eased only slightly. Both figures still sit far below OpenAI’s reported $40B and Anthropic’s $65B annualized run rates, but they are the first sign that the distillation accusations and federal advisory this thread tracks have not visibly dented Moonshot’s own growth trajectory or K3’s usage. (TechCrunch)

2026-09-11 — Anthropic’s own threat report names four more Chinese labs running silent-relay schemes against Claude, with exact exchange counts and a technical bypass of Anthropic’s chain-of-thought protection

• Anthropic’s September threat-intelligence report names DeepSeek, Moonshot, Zhipu (Z.ai) and Xiaomi as running distillation and data-misuse operations against Claude beyond the Qwen/Alibaba case already on this thread, and for three of them documents something worse than distillation: silently routing their own paying customers’ live requests to Claude, serving Claude’s answers back under their own brand, and harvesting the exchange for training data, without telling the customer. Per Anthropic’s own document: Moonshot (GTG-16002) relayed roughly 300,000 Kimi-branded customer requests to Claude Opus in a ten-day sample (23M+ exchanges attributed May–July). DeepSeek (GTG-16001) ran the same silent-relay scheme, 12.1M+ exchanges over fourteen days in July. Zhipu/Z.ai (GTG-16006) ran a chain-of-thought extraction pipeline against Claude Opus 4.8 (770,609 exchanges over ten days in June) to clean GLM training data, then separately targeted the cyber capabilities of Anthropic’s Fable model and another leading US lab’s top model ahead of GLM 5.3, giving up on Fable once its safeguards degraded the attack. Xiaomi (GTG-16008) replayed its own MiMo customer sessions through Claude to generate training data without serving the answers back. Both Moonshot and DeepSeek are reported to have defeated a specific Anthropic technical control — the “thinking signature” token that stands in for raw chain-of-thought in the API — via a cross-session replay trick that induces Claude to reconstruct its own reasoning trace from the signature, a materially more sophisticated evasion than the prompt-injection method used against the Qwen models already on this thread. (Anthropic)

2026-09-10 — Moonshot AI is now exploring a dual Hong Kong + Shanghai STAR Market listing, on top of its existing confidential HK IPO filing

• The South China Morning Post reported Moonshot is discussing a dual listing — its already-known confidential Hong Kong IPO filing (Reuters-reported target ~$3bn) plus a subsequent mainland listing on Shanghai’s STAR Market — with sources citing weaker recent performance of Hong Kong-listed AI stocks and a crowded HK IPO pipeline as reasons to hedge onto STAR, which has recently drawn other Chinese tech listings (e.g. Unitree Robotics). This is a capital-structure update to the same IPO thread that logged the $3-5bn HK target revision and a $50bn pre-money private round on 09-04, arriving one day after Moonshot was named by name in the CISA/NSA/FBI distillation advisory this thread already recorded on 09-09 — worth watching whether the advisory affects listing terms or investor appetite. Moonshot did not comment. (DealStreetAsia, AOL/SCMP wire)
• Anthropic’s own newly published threat-intelligence report names, for the first time, the specific downstream models trained on harvested Claude reasoning — Qwen 3.5, 3.6 and 3.7 — escalating the distillation dispute this thread tracks from volume claims to named products. Anthropic’s “Detecting and countering misuse of AI: September 2026” report (internal case designator GTG 16005) attributes the campaign to Alibaba’s Qwen/Tongyi Lab, not Moonshot — the largest chain-of-thought distillation campaign Anthropic says it has ever measured against Claude, peaking at nearly 3 million exchanges per day from more than 3,500 fraudulent accounts, using prompt injection to force Claude Opus 4.6 and 4.7 to write out their reasoning before answering. This is a different Chinese lab than the one this thread is named for, but the same six-company federal distillation advisory (CISA/NSA/FBI, 09-09, logged below) that named Moonshot also named Alibaba — so today’s report gives fresh, named-model detail to one of that advisory’s other members. It is also distinct from Anthropic’s 2026-06-10 letter to the Senate Banking Committee on the same Alibaba dispute, which cited 28.8 million exchanges across 25,000 accounts for April–June and did not name the trained models; today’s report gives a higher daily rate and names Qwen 3.5/3.6/3.7 for the first time. (Anthropic)

2026-09-09 — CISA, NSA and FBI jointly name Moonshot and five other Chinese labs in a formal federal distillation advisory

• CISA, the NSA and the FBI published a joint advisory (AA26-251A) formally naming Moonshot AI — alongside DeepSeek, Alibaba, MiniMax, StepFun and Z.AI — as running “industrial-scale,” “aggressive, malicious, and targeted” knowledge-distillation campaigns against Claude, GPT, Gemini and Grok since late 2024, using APIs, cloud resellers and proxies to evade access restrictions, “likely with Chinese government awareness.” This escalates the accusation this thread has tracked since Dario Amodei’s informal one against Moonshot specifically in June into a formal, six-company federal attribution naming Moonshot by name. China’s foreign ministry (Mao Ning) called it “unfounded” the same day. (CISA, Bloomberg, NBC News, The Register, Washington Times)

2026-09-04 — Moonshot’s Hong Kong IPO target revised up to $3-5bn, new $50bn pre-money round surfaces

• Bloomberg reports Moonshot is now considering raising up to $5 billion (a $3-5bn range) in its Hong Kong IPO, up from the roughly $3bn target reported on 09-03, and has added Bank of America as overall coordinator alongside the previously-reported CICC, Deutsche Bank and Goldman Sachs. The same report surfaces a parallel, separate development: Moonshot is reaching out to backers for a new pre-IPO funding round at a $50 billion pre-money valuation — a step up from the $35bn valuation set by its August Series F ($3.5bn raised). Bloomberg’s own framing stresses “considerations are ongoing and details such as size and timing could change.” (Bloomberg via Investing.com, Bloomberg — 403’d direct fetch, corroborated via mirror)

2026-09-03 — Moonshot reportedly files confidentially for a Hong Kong IPO targeting $3B at a $50B valuation

• Moonshot AI has reportedly submitted a confidential A1 application to the Hong Kong Stock Exchange this week, formally starting its IPO process at a target raise of about $3 billion against the ~$50 billion valuation this map has been tracking as a private-round target. Goldman Sachs, CICC and Deutsche Bank are named as the banks on the deal; Moonshot had to unwind its offshore incorporation structure into an onshore China domicile before filing, per the reporting. The company’s only comment: it has “no information to disclose.” This lands well inside the map’s own moonshot-hk-ipo-filing ledger window (due 2026-09-30, “as soon as this month” per the original August sourcing) and effectively answers it early, pending confirmation on HKEX’s own disclosure. (TechNode, RTE/Reuters, Invezz)

2026-09-01 — Anthropic ships an anti-distillation API restriction and, for the first time, quantifies the campaign it says targets its own models

• Anthropic disclosed it traced more than 16 million Claude exchanges to distillation campaigns run through roughly 24,000 fake accounts, and named three Chinese labs — DeepSeek, Moonshot AI and MiniMax — as beneficiaries, alongside the Fable 5.1/Mythos 5.1 launch. This is the first time Anthropic itself has put a number on the distillation activity behind the OSTP accusation this thread has tracked against Moonshot specifically since 07-23; Moonshot is now one of three labs named rather than the sole one. (Anthropic, Yahoo Tech)
• Anthropic shipped a concrete technical countermeasure: API accounts opened on or after 2026-08-31 can no longer carry Claude’s extended-thinking blocks forward into a manually edited prior-context conversation — a technique Anthropic says “has also been used to distill its models at scale.” Accounts opened before that date are exempt, and ordinary Claude Code/Cowork/claude.ai users see no change; this is the first shipped product response on this thread’s own record, as opposed to accusation, denial, or trade-policy threat. (Anthropic, Yahoo Tech)

2026-08-25 — Moonshot seeks US cloud distribution for Kimi K3

• Moonshot AI opened talks with Microsoft, Amazon and Google to host Kimi K3 under a revenue-share worth up to 30% to the clouds. Reuters reported the discussions as early, with the split, data access and token-tracking mechanics all unresolved. It inverts the usual direction of this story: a Chinese lab selling through US infrastructure rather than building away from it. (Reuters, via Yahoo Finance)

2026-08-19 — The Moonshot allegation broadens: three more Chinese labs, and a bill to close the loophole

• CNBC reports Moonshot, ByteDance, Alibaba, and Tencent have all been accessing restricted Nvidia compute via data centers in Thailand, Malaysia, and Japan — legal today because US export controls govern physical chip transfer, not remote cloud access. This extends the thread’s 07-23 entry (which alleged Moonshot specifically obtained GB300 access via Thailand) to three more major Chinese AI firms, and names the legislative response: the Remote Access Security Act (RASA), which passed the House in January 2026 and is pending in the Senate — it would extend export-control authority to remote/cloud access, though a separate rulemaking would still be needed to apply it to chips. (CNBC)

2026-08-15 — Beijing, not Washington, is what actually derails Moonshot’s IPO clock

• Moonshot converted from a limited-liability company to a joint-stock company in early August and is unwinding the offshore “red-chip” entity it used to raise foreign capital, after Beijing tightened rules barring strategic-tech firms with overseas holding structures from listing abroad. Reported by the Financial Times (08-08); officials reportedly hardened the policy after Meta’s acquisition of Manus AI raised concern about foreign influence over domestic AI firms' governance. Insiders now say a Hong Kong listing is unlikely before 2027 unless the restructuring — either overseas investors standing up onshore vehicles, or selling out and rebuying onshore — clears quickly. (Seoul Economic Daily, citing FT, National Technology)
• This reverses the trajectory this thread has tracked since 08-03/08-04 — Moonshot was chasing a ~$50B Hong Kong listing “as soon as this month” and had opened final pre-IPO talks; the same listing is now pushed out at least a year. Notably, neither source ties the delay to the Kratsios/Bessent distillation accusation or an Entity List threat — this is Beijing’s own capital-control anxiety over foreign ownership of strategic AI IP, a second-order consequence of the same “who controls China’s frontier AI” question this thread’s Watch line is built around, running on a different axis than the US side of the dispute. Moonshot reportedly has ample funding (~$30-50B valuation across two rounds, the second backed by Chinese state funds) to wait out the delay without urgency.

2026-08-04 — Moonshot opens final pre-IPO talks

• Opened final pre-IPO funding talks, targeting up to $50B pre-money ahead of the ~6-month Hong Kong listing already tracked, riding reported Kimi K3 benchmark strength against GPT-5.5/Claude Opus 4.8/GLM-5.2. ⚠ Loosely dated (“opening in August”).

2026-08-03 — Moonshot reportedly chases a ~$50B Hong Kong IPO amid the Fable-distillation dispute

• Moonshot is said to be targeting a ~$50B Hong Kong IPO “as soon as this month” — above its $35B Series F and sitting inside the OSTP allegation (07-22) that it distilled Kimi K3 from Anthropic’s Fable via covertly-owned GB300 servers routed through Thailand; the figure is single-source-thin, citing unnamed “reports coming out of China.” (Wccftech)

2026-07-27 — Beijing draws the line

• China vowed “all necessary measures” against the threatened US sanctions on AI firms over the Kimi-K3-distilled-from-Fable-5 accusations — tariff lines drawn ahead of the Sept 24 Xi–Trump summit. The state response the Watch line waits on has begun, on the trade axis rather than the export-control one. (Bloomberg)

2026-07-26 — K3 weights live

• Kimi K3 open weights shipped a day EARLY — billed the largest open-weight model ever, mid distillation-accusation and mid $50B pre-IPO push; David Sacks publicly pushed back on the “Chinese model” framing the same day.

2026-07-23 — The White House names a name

• OSTP Director Michael Kratsios directly accused Moonshot of distilling Anthropic’s Fable model to build Kimi K3 — the first time a senior US official has named a specific Chinese lab; separately alleged Moonshot obtained restricted Nvidia GB300 chips via Thailand. Treasury Secretary Bessent: “sanctions and the Entity List are both on the table.” Under Secretary of State Helberg called it a “heist of invaluable American IP.” (SCMP)
• Moonshot denies it — only 15 days elapsed between Fable’s July 1 release and K3’s July 15 launch; independent researchers (Redwood’s Ryan Greenblatt) call the evidence “thinner than official statements suggest,” circumstantial (K3 self-identifying as Claude at unusual rates has innocent explanations — scraped transcripts, leaked system prompts).

← Backstory

(Split out of China’s Stack 2026-07-24 — see that thread’s 07-17→07-22 entries for the Kimi K3 launch/chip-selloff backstory this accusation follows.)