The Projection — a symmetric watercolor butterfly

The Projection

The surface is never the system.

AI

Distillation Fight

The White House (OSTP's Kratsios) directly accused Moonshot of distilling Anthropic's Fable model to build Kimi K3, plus obtaining restricted Nvidia GB300 chips via Thailand; Bessent named sanctions/Entity List as options. Moonshot denies it; independent researchers call the evidence circumstantial. Lands right before K3's open weights ship (07-27). Watch: any formal Commerce/Entity List action, Moonshot's response, independent verification one way or the other, and whether it chills enterprise adoption of K3.

STATUS · OPEN OPENED · 2026-07-24 LAST SEEN · 2026-09-04
Moonshot AI Anthropic

Summary

China vowed "all necessary measures" against threatened US sanctions on AI firms just as Kimi K3's open weights shipped a day early.

Whether Washington follows through with formal Commerce or Entity List action against Moonshot over the distillation and chip-smuggling accusations remains open.

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

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

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

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

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

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

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

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

2026-07-27 — Beijing draws the line

2026-07-26 — K3 weights live

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

← 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.)

This week's evidence

🕓 GLM-5.5 has not shipped, and the afternoon check suggests the name itself may be the problem with this expectation. Z.AI's Hugging Face org (zai-org) carries GLM-5.3 and its Flash variants — with checkpoint uploads hours old, so the channel is demonstrably live and publishing something else — and no GLM-5.5 model card exists anywhere. Chinese tech coverage says Zhipu's numbering stayed on the 5.x track and that "5.5" was a rumoured designation the company never used. ⚠️ This does not mean the expectation is satisfied by GLM-5.3. The ledger claim is a >1T-parameter, 1M-context model; GLM-5.3, which shipped 08-14 and is already on this map, is 753B. So the honest position is that the claimed model does not exist under any name yet, and the label it was logged under appears to have been abandoned by the vendor. 2026-08-31
Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on 09-01 — one model at two safeguard levels, Fable generally available and Mythos restricted to vetted cyber and life-sciences users through two trusted access programs, the latter "developed in partnership with the US government." Anthropic's own numbers: Terminal-Bench-Science 52.6% vs Fable 5's 24.7% (Opus 5 29.0%, GPT-5.6 Sol 22.4%); Terminal-Bench 4.0 55.8% (Mythos 5.1: 60.9%) vs 42.0%; AutomationBench 31.4% vs 17.1%; Humanity's Last Exam 60.9% without tools. Artificial Analysis puts it at a record 66 on its Intelligence Index, at 20% higher cost per task because it writes ~1.7x the tokens. Cache reads cut 75% to $0.25 per million tokens — ~25% cheaper on typical workloads, up to ~45% on agentic ones; standard rates unchanged at $10/$50. New in this release: default text watermarking under the EU AI Act code of practice with a detection API in private preview; "Enterprise Frontier Safeguards," a phased customer-held zero-data-retention option; cyber safeguards that intervene ~60% less and now allow vulnerability discovery (not exploit development); and strengthened anti-distillation mechanisms, with Anthropic disclosing for the first time that it traced 16 million exchanges to ~24,000 distillation accounts and naming DeepSeek, Moonshot and MiniMax. Neither the announcement nor the system card names CAISI, the UK AISI or the 30-day review framework. ⚠️ Missed for two days with 45-50 hits in this map's own buffer; recorded on Distillation Fight and Frontier Gatekeeping where it bears on a watch line, and offered as a thread candidate. (Anthropic, system card, Artificial Analysis, VentureBeat, TechCrunch) 2026-09-01
Moonshot AI filed confidentially for a Hong Kong IPO this week — an A1 application targeting about $3bn at a roughly $50bn valuation, with Goldman Sachs, CICC and Deutsche Bank on the deal — two days after Anthropic named it, with DeepSeek and MiniMax, as a beneficiary of the distillation campaigns it traced to 16 million exchanges and ~24,000 fake accounts. Moonshot had to unwind its offshore structure into an onshore domicile to file; its only comment was "no information to disclose." This resolves the ledger's 08-03 filing expectation 27 days early; the separate ~$50bn private round never closed on the record and stands passed-silent. (Reuters via RTÉ, TechNode) 2026-09-03
Moonshot AI filed confidentially for a Hong Kong IPO targeting ~$3bn at a ~$50bn valuation (Goldman, CICC, Deutsche), per Reuters and TechNode. The second frontier-lab listing process on this map after SpaceX's, and the first from China's stack; the 08-03 filing expectation resolves 27 days early, while the private ~$50bn round never closed on the record. (Reuters via RTÉ) 2026-09-03
Moonshot is now considering raising $3-5 billion in its Hong Kong IPO, up from the roughly $3 billion reported on 09-03, has added Bank of America as an overall coordinator alongside CICC, Deutsche Bank and Goldman Sachs, and is separately sounding backers on a pre-IPO round at a $50 billion pre-money valuation — a step up from the $35 billion of its August Series F. Bloomberg's own framing: "considerations are ongoing and details such as size and timing could change." Reported at 13:52 ET, just before this window opened, and not on the record until now. (Bloomberg, via Investing.com) 2026-09-04

Related threads (shared entities)

· Microsoft's Hedge
· Circular Financing
· Frontier Gatekeeping
· China's Stack
· In-House Silicon
· Lab IPO Wave
· Anthropic IPO
· Allianz AI Claims
· The #2 Cashes In
· The Rogue Agent
· Copyright Exposure
· Anthropic Rents the Buildout
· The Backlash Prices In
· The Enterprise Agent Land Grab