AI Morning Brief · Nov 16, 2026
Top Stories
1.NVIDIA CEO drama escalates: Huang reportedly weighs a comeback as allies lobby the board; Bessent says pick an outsider and explain any suspect China chip sales within 10 days
The FT reported on Nov 15 that Jensen Huang is considering returning as NVIDIA CEO and has discussed it with confidants, while allies are assembling a Silicon Valley executive "war room" to lobby the board and seek White House support; Elon Musk and Epic CEO Tim Sweeney have publicly backed a "Bring Jensen Back" petition, and major shareholders Brad Gerstner (Altimeter) and Gavin Baker (Atreides) wrote to the board on Nov 13 asking that Huang's return not be ruled out. Treasury Secretary Scott Bessent said on Fox News Sunday on Nov 15 that the board should pick a successor from outside the company, that a "major overhaul" is needed, and that the board must explain within 10 days (about Nov 25) whether the company pushed suspected illegal chip sales to China. NVIDIA and Huang both declined to comment, and there is no announcement of the board agreeing to a return. (FT original via multiple relays; Bessent's remarks via the NYPost and ShiftDelete citing the same Fox interview)
Sources: tradingview.com · cryptonews.net · nypost.com · bulletin.observer
2.Anthropic IPO follow-up: up to 20% reserved for retail investors, about 165M shares; about six-times oversubscribed; pricing Nov 18, trading Nov 19
The Wall Street Journal reported on Nov 15 that Anthropic is reserving up to about 20% of its Nasdaq IPO (about 165M of 825M shares) for individual investors, roughly double the usual share; the price range is $70–78, pricing is set for Nov 18 and trading starts Nov 19 under ticker ANT, with a valuation of up to about $1.4T; retail channels include Fidelity, Schwab, Robinhood and SoFi. Bloomberg, citing people familiar with the matter, says orders are about six times the shares offered; the amended S-1 discloses compute and infrastructure commitments of about $895B (another reading: $900.9B), with Google and NVIDIA investing about $10B each as part of a $26B private placement from seven strategic partners. (WSJ, CNBC and Bloomberg accounts align; this resolves the previous issue's unverified items on the six-times oversubscription, the Nov 18 pricing and Nov 19 trading dates, and the $10B each from Google and NVIDIA)
Sources: at-pulse.com · at-pulse.com · systemdesign.academy
3.Meituan releases LongCat-2.5 weights: a 1.6T-parameter agent model trained entirely on Chinese hardware, scoring 88.4 on Terminal-Bench
Meituan released the full weights of LongCat-2.5 on Nov 15 under the MIT license (commercial use allowed), one week after the model was officially unveiled around Nov 9; the MoE model has 1.6T total parameters with about 48B active per token, and the company says it was trained entirely on Chinese hardware (the predecessor LongCat-2.0 was trained on a cluster of more than 50,000 Chinese AI chips), with a sample deployment of 194 Huawei Atlas A2 accelerators (2 for image processing, 64 for prompt ingestion, 128 for response generation) and no guidance for NVIDIA GPUs. Meituan reports 88.4 points on Terminal-Bench v2.1, slightly below Claude Opus 5 (89.1) and GPT-5.6 Sol (88.8), with DeepSeek-V4.1-Flash on top at 90.6 (all vendor-reported, with no independent reproduction yet); the API is priced at a limited-time promotional rate of about €0.27 per million input tokens and €1.07 per million output tokens. (The Decoder, Nov 15, cross-checked across sources; the 1.6T versus 1.8T Hugging Face metadata discrepancy remains unverified, see Unverified Items)
Sources: the-decoder.com · dev.to · agibytes.net
4.ByteDance opens the Seedance 2.5 Flash video-model API at 14:00 on Nov 15
The video model, already live in products such as Jimeng since Nov 12, opened for API calls and online trials on Volcano Engine at 14:00 Beijing time on Nov 15, with BytePlus ModelArk opening overseas in parallel; the company says generation is about twice as fast and 720p is about 34% cheaper than the flagship (about 0.99 yuan per second on Volcano Engine, about $0.152 per second on ModelArk overseas), but 1080p and 4K are not supported, and the "nearly twice as fast" claim has no public benchmark to back it yet. (Multiple sources; the speed claim is the vendor's)
Sources: allagent.wiki · mlllm.io · startupfortune.com
5.Cursor's full disclosure (follow-up): the malicious Claude Code extension came from a Fireworks AI leaked key, and up to 5,000 users may have received it
Cursor disclosed in full on Nov 13: attackers used an access key Cursor had sent to Fireworks to copy a bucket of historical coding-agent trajectories, found a small number of unredacted credentials in it, and on Oct 29 planted tampered Claude Code extensions 2.1.312 and 2.1.313 in the Cursor extension marketplace cache; up to 5,000 users may have received them within about 11 hours; fewer than 0.01% of customers' API credentials and fewer than 0.05% of users' trajectories were affected; the last attack activity was seen on Nov 1 and has been contained; Cursor has moved all inference off Fireworks. (Cursor's official update as the primary source, followed by BleepingComputer on Nov 14 and others; the previous issue only reported that Cursor confirmed attackers had obtained one of its keys — this is the full disclosure)
Sources: aiworldwire.org · windowsforum.com
6.CNN Megaspeed investigation (follow-up): in three years it became one of Asia's biggest NVIDIA chip buyers, with over $4B in purchases; Megaspeed says Alice Huang left on Aug 30, 2025; Bessent blames Lutnick for delaying the Entity List listing
A CNN investigation published Nov 14 found Megaspeed went from an inactive Chinese gaming subsidiary to one of Asia's largest buyers of NVIDIA chips within three years, with cumulative purchases of more than $4B; in a new statement on Nov 15 the company said Alice Huang left on Aug 30, 2025, holds no current role, and called the media claims unsupported by "verifiable evidence". CNN reported on Nov 15 that Bessent blames Commerce Secretary Lutnick for delaying putting Megaspeed on the Entity List, even though the Treasury, State Department, Pentagon and USTR have supported listing it for months; the White House said on Nov 13 that it still fully trusts Lutnick, and Lutnick is still expected to travel with Trump to China next week. Separately, Megaspeed CEO Tan Yong Pong said on Nov 13 that the company has cooperated with Commerce Department checks and hired a US law firm for an independent compliance review; Singapore's Home Affairs Ministry says there is insufficient evidence that Alice Huang committed related offenses, but barred her from re-entering Singapore for giving false information that downplayed her role; Singapore police are still investigating Megaspeed. (CNN original investigation plus relays; the core "diversion to China" allegation remains "alleged" with no conviction; Bessent's blame of Lutnick is a single-source claim from anonymous officials, see Unverified Items)
Sources: cnn.com · demokratis.co.id · cnn.com
7.BAAJA follow-up: the Senate is set to hold its first cloture vote on Monday evening, Nov 16 ET, with a second at about 01:00 on Nov 18
The Senate is scheduled to hold its first cloture vote on the amended BAAJA text at about 17:45–18:00 ET on Monday, Nov 16 (60 votes needed), with the second at about 01:00 on Nov 18; the votes themselves fall outside the window — what is new in the window is the confirmation of the schedule and the clause details. The bill makes data centers bear all new grid costs (§2107, including the unrecovered cost of substations already built even after a data center closes, protecting residential bills); 17 Democratic state attorneys general wrote on Nov 9 opposing it, saying it weakens state and tribal sovereignty. (Hill Heat and Kunkafa agree on the schedule; this resolves the previous issue's unverified item on the two cloture times, which rested on Hill Heat alone)
Sources: kunkafa.com · hillheat.news
8.Meta's Muse expands to Apple Watch and Wear OS smartwatches
Muse now supports voice calls, texting and dictation from the wrist; the iPad app launched a month earlier; availability is limited for now to the US, Canada, Japan and Singapore. (TechCrunch original plus multiple relays)
Sources: cryptoversenews.eu · thesbharat.com
9.Tencent's "50% layoffs" rumor debunked (background, Nov 12, first recorded in this issue)
Tencent PR director Zhang Jun responded on Weibo on Nov 12 to the "50% layoffs" rumor, calling it "pure misunderstanding and a rumor" and "completely untrue"; the source was a "half army, half navy" metaphor from management at the Nov 11 Tencent 28th-anniversary Linktime livestream (50% of people defending the core business, 50% exploring AI-native work), which was misread as a layoff ratio; people close to Tencent told The Paper, Yicai and Securities Times that the claim is untrue. Tencent has published no specific headcount adjustment plan, ratio or timetable. (ITHome original, Nov 12, via multiple relays; this resolves the previous issue's unverified item on Tencent cutting about 50% of Yuanbao and Hunyuan staff)
Sources: aitop.news · finance.china.com
Platform & Model Tracking
OpenAI (GPT/ChatGPT)
Single source: OpenAI has disbanded its independent robotics team for the second time, with researchers moving to core research and application and hardware teams — the company says the move is meant to have general frontier models control robots directly, in line with a strategic shift; the daily digest is the only source, with no official announcement or mainstream corroboration (unverified). GPT-6.2 remains unreleased — as of a Nov 13 check, OpenAI's official docs, API changelog, pricing pages and newsroom contain no GPT-6.2 entry; "gpt-6.2-astra" appears only in individual users' Codex usage screenshots, while the model picker still shows GPT-6 Astra; a Nov 13 post claiming a delay "because of Claude Fable 5.2" has been debunked; Nov 6 (the earliest Friday cited by The Information) has passed with no release. No new ChatGPT Wallet developments in the window (after its Nov 13 launch, only analysis and commentary appeared, such as a Visa trust index finding only 23% of US consumers trust GenAI with payments). Single source: the FT reported on Nov 15 that Samsung SDS is in early talks with Anthropic and OpenAI to take them on as data-center tenants (a relay, early stage, no terms or capacity figures).
Sources: news.poweryourprocess.ai · orcarouter.ai · fourweekmba.com
Anthropic (Claude)
IPO retail allocation and oversubscription (see Top Story 2). Single source: Anthropic launched a Claude Meetings beta that can join Google Meet calls for live collaboration — for paid Team and Enterprise plans, Claude joins video meetings as a participant to answer questions live and co-write project proposals and schedules; external meetings are opted out by default and internal data queries need prior approval (unverified). Single source: Claude Opus 5.5 has become the default model for all paid plans, with a prompting guide published (unverified). Single source: TSMC's board approved two investments of up to $500M each to take part in Anthropic's share subscription (unverified; consistent with TSMC being one of the seven private-placement parties, but the amounts have no second source). S-1 compute-commitment figures: the amended S-1 discloses about $895B in compute and infrastructure commitments (another reading: $900.9B), about 11x its $81B annualized revenue — broadly consistent with the previous issue's "$900.9B cumulatively through 2045".
Sources: at-pulse.com · at-pulse.com · kimkj.com · korben.info
Google DeepMind (Gemini)
Single source: Gemini 4 Carbon topped the MathArena mathematics benchmark, replacing GPT-6.1 Sol at the top alongside Claude Fable 5.2 (unverified). No other new release in the window; Gemini 4 Hydrogen entering Search was Nov 12 news, already covered.
Sources: news.poweryourprocess.ai
Meta
Muse expands to smartwatches (see Top Story 8). Single source: HUMAN Security's October "State of Agent Traffic" report says agents from Meta and OpenAI together account for nearly 90% of agent requests on the sites it protects, with Muse alone at 59%; total agent requests in October were 8x August's and about 2.6x September's; Muse's request volume nearly quadrupled within a month and is 3.7x ChatGPT Agent's; Muse also moved its Web Bot Auth signing key from meta.com to its own muse.ai domain on Nov 7–9, with about half of requests unsigned as of Nov 10 (a single vendor's report, unverified). Single source: Meta is internally testing a Llama-powered prediction-market app (previously codenamed Arena, Antwerp and FBForecast), with Llama writing questions on trending topics, recommending markets and settling them in near real time, aimed at users aged 18–34, with iPhone and Android versions planned after employee testing (unverified). Single source: Meta released the Muse Ember inference model (for repetitive agent work) and Muse Wonder (planning) — Ember targets coding subtasks, running tests, targeted fixes and long-document parsing, with adjustable reasoning strength; parameter counts were not disclosed (a LinkedIn newsletter, reported Nov 15, with the actual release date not stated, unverified). The privacy researcher Paul Walsh's Muse profiling allegation and the Muse Chrome extension have no Meta response or formal announcement in the window (covered in the previous issue, see Unverified Items).
Sources: cryptoversenews.eu · paktechstan.com · aibriefing.dev · linkedin.com
xAI (Grok)
Grok 4.8 still has no official release — Musk said on Nov 10 that pretraining was complete, RL had begun and a soft release to some Grok Bot users was planned later in the week; social posts on Nov 12 claimed a quiet rollout to a small group of Bot users, but xAI's news page, model catalog and API show no 4.8 record, with no model ID, pricing or benchmarks; the 2.5T-parameter and new C++ training and inference stack (dropping JAX and SGLang) claims still rest only on Musk's own posts and tracking pages. No Grok 5 release or pricing change in the window.
Sources: explainx.ai · grokshelf.com
Microsoft
No significant new developments in the window (Copilot or Azure AI). Microsoft Ignite runs Nov 17–20. The five-year Azure partnership with Ohio State University (about 70–80% of compute moving to Azure, reported at $250M) was announced on Nov 10, outside the window.
Sources: gate.com
NVIDIA
CEO drama and Bessent's remarks (see Top Story 1). Q3 earnings are due after the close on Tuesday, Nov 17: guidance is about $108B in revenue (±2%), with the market consensus at about $108.8–109.2B and EPS of about $2.47–2.48, both up about 90% year over year; gross-margin guidance is about 74%; interim CEO and CFO Colette Kress hosts the call (17:00 ET); the stock closed at $186.64 on Nov 6, about 22% below its Oct 6 record high. Correction: the House's written-confirmation deadline for the Dec 1 closed-door interview was 20:00 ET on Friday, Nov 13 — the previous issue's main text misstated it as Nov 14 (its Unverified Items entry said Nov 13, so the two were inconsistent; this issue corrects it to Nov 13 per the letter); as of the evening of Nov 14 (20:30 ET update) there is no public report of Huang's written confirmation or of a subpoena being issued; House Foreign Affairs chair Mast said on Nov 12 that Huang had agreed to testify, and China Committee chair Moolenaar said on Nov 15 that Huang had agreed with the date still being negotiated — both claims conflict with the "no public response as of Saturday evening" reporting (see Unverified Items). Commerce Secretary Lutnick has agreed to appear before the Senate Banking Committee on Dec 10 (CBS, via multiple relays). The CEO search is still at an early stage: as of Friday, Nov 13, the board said it was still selecting a search firm and had "not formally approached any candidate", calling named speculation "premature" (confirmed again by Nov 15 reporting as the latest board position; the FT reported on Nov 13 that the board had approached Arm CEO Rene Haas, which the board has neither confirmed nor denied for any specific contact, and Reuters says it could not independently verify it).
Sources: tradingview.com · nypost.com · smwmarkets.com · agenccy.ai · sovereignmagazine.com · aiagentupdate.com
Mistral
No significant new developments in the window. The official news page's latest items as of Nov 14 are still the Nov 5 Studio expansion and the Oct 29 Large 4 open weights.
Perplexity
The Teammates Mac launch was Nov 13 news (Max subscribers only, $200/month), and Nov 15 reports are re-reporting, so it is not repeated. Decider v1.3 earned Artificial Analysis scores (follow-up): released Nov 12; on Nov 14 the Artificial Analysis Decision Index v2.1 scored v1.3 at 70 points (6th) and Flash v1.3 at 64 points (10th, up 4 from v1.2); Apache 2.0 open source, 256K context, hosted at $0.02/$0.01 per million tokens.
Sources: alphasignal.ai · explainx.ai
China AI Companies
Meituan (LongCat)
LongCat-2.5 weights released (see Top Story 3). The parameter-count discrepancy — 1.6T per Meituan versus 1.8T suggested by Hugging Face metadata — remains unverified (see Unverified Items).
Sources: the-decoder.com
ByteDance (Doubao)
Seedance 2.5 Flash API opened (see Top Story 4). Zhang Yiming's remark that ByteDance cannot catch up with OpenAI and Anthropic until 2029–2030 is Nov 12 news, with no follow-up in the window. No new developments on the reported resignation of AI chip chief Shi Yunfeng (see Unverified Items).
Sources: allagent.wiki · mlllm.io
Zhipu (GLM) / MiniMax
AutoClaw quota rotation — the 50M tokens claimed by logging in Nov 9–15 (100M for the overseas version) expired at 23:59 on Nov 15; the new round runs 00:00 Nov 16 to 23:59 Nov 22, granting 100M tokens (10,000 points) per login, one claim per account, with the two rounds not stacking and no early claims; the quota is in-client only and cannot be exported to the API (two quota-tracking sites agree, both third-party summaries rather than Zhipu's official announcement). A ZCode weekend pack of 500M GLM-5.3-Flash tokens expires at 09:00 on Nov 16, after the window's 07:00 cutoff — noted for tomorrow's issue, not counted in this window. Hong Kong stocks: no trading on Sunday, Nov 15; the Nov 13 closes were covered in the previous issue.
Sources: token-fbi.com · aihr.wang
Tencent (Hunyuan/Yuanbao)
The "50% layoffs" rumor was debunked (see Top Story 9). The free access to Hunyuan Hy4 and Hy3 in clients ended at 23:59 on Nov 15 — the daily free quota for Hy4 and Hy3 inside WorkBuddy and CodeBuddy ended then, with calls charged against points again afterwards; two other offers are unaffected: a 2.5x discount on Hy4 points under the Token Plan subscription until Nov 17 or 18 (sources differ) and a 50% discount on TokenHub pay-as-you-go list prices until Dec 31. Hy4 weights still not released as of Nov 15.
Sources: aitop.news · allagent.wiki · aihr.wang
DeepSeek
Consumer quota-management check (Nov 14, the night before the window): a check of the Nov 13 live build of chat.deepseek.com shows quota metering, exhaustion prompts and five-minute polling have reached the chat page — but across 603 JS chunks there are zero hits for "Coding Plan", "plan", "subscription", "membership", "purchase" or "payment", and the official pricing page still lists only per-token API pricing with no subscription items; the inference by Kuai Keji (Nov 13) and ITHome (Nov 14) that a paid plan is coming is media speculation, and the company has said nothing. Conclusion: consumer users may soon hit a usage cap for the first time, but whether the cap is followed by waiting or by paying to unlock has no official answer (an update to the previous issue's Unverified Item 3, still unverified). No new model release in the window; the second-round financing figures of about 100B yuan versus 105.5B yuan remain unconfirmed, with no official statement (see Unverified Items). The former OpenAI engineer's claim that DeepSeek-V4 and five other Chinese models carry NVIDIA Blackwell initialization "fingerprints" saw Korean- and English-language second-hand readings on Nov 15 — window chatter, not new facts (see Unverified Items).
Sources: seoml.com · seoml.com · zero0ne.dev · amjadphd.blogspot.com
Alibaba (Qwen)
The Qwen4-Max preview (Nov 11 to Nov 25) brought no new version, pricing or benchmark in the window. Claims of four Qwen4 versions and more than 3T parameters remain unverified (see Unverified Items).
Baidu (ERNIE)
No new developments in the window; ERNIE 5.5 Flash Preview (Nov 11) and Baidu World 2026 (Nov 12) are both pre-window news.
Moonshot (Kimi)
Kimi 3.1 Pro was featured in GitHub's weekly open-source picks (ThinkAI's Nov 15 weekly lists it first, saying the full weights are in the official model repository, with 2.8T total parameters, about 1M-token context and text-plus-image input) — a retrospective mention of a weights release from early November, with no new model action in the window. Reports of Kimi K4 (Q2 2027, 5.4T parameters) and K3.5 are Nov 7 news, with no new developments in the window.
Sources: thinkai.news
Manus
Single source: Manus was reported to run paid promotions pitching an "easy side hustle", with Meta saying it was unaware and is reviewing — The Verge reported that Manus, acquired by Meta last year for $2B, paid third-party agencies for UGC on TikTok, Instagram and YouTube claiming "10 minutes to start, $5,000 a month"; a Manus spokesperson acknowledged working with third-party agencies on paid promotion but denied endorsing exaggerated earnings claims and said the content is under review; Meta did not respond on whether it knew; the report also notes the acquisition has been halted by Chinese regulators and Meta is seeking a solution with Beijing. The link is a mirror or aggregator of the Verge report, with no second independent source found (single source, unverified, see Unverified Items). No new Manus product or model release in the window.
Sources: thevergetoday.pages.dev
Other China developments
China's AI export-control draft comment period closed Nov 15 with no new action in the window — the Ministry of Commerce and Ministry of Science and Technology's Announcement No. 48 proposes adding 40 technologies (23 of them AI), with the comment period running Oct 16 to Nov 15; on Nov 15 only updates and re-posts of existing content appeared (such as Asia AI Policy Monitor issue 52 restating the Nov 15 deadline), with no official release of a final list, extension or summary of comments; the interim pre-export registration requirement has been in effect since Oct 16.
Sources: aicoder.com · asiaaipolicymonitor.substack.com
Policy and regulation
BAAJA (see Top Story 7). Chip export-control bills in the FY2027 NDAA (follow-up, no new text): Senate Majority Leader Thune and Armed Services chair Wicker told Semafor on Nov 12 that keeping the three chip bills in the final NDAA text is "not yet agreed" but that stricter controls are "highly likely" — the statement itself is from Nov 12, and what is new in the window is Nov 15 re-posting and synthesis (a single-source relay, with no link to the Semafor original); the three bills are in the Senate NDAA manager's package (the AI Overwatch Act, the Chip Security Act and the MATCH Act), while the fourth, the Remote Access Security Act, is not in the Senate defense text; House Speaker Johnson said on Nov 12 that Congress is forming a "developing consensus". Treasury Secretary Bessent said in a Nov 13 CNBC interview that the government already has enough export-control authority and the key is enforcing existing tools rather than new legislation (a Nov 15 relay, single-source, not independently verified). Megaspeed (see Top Story 6). EU AI Act: no new developments in the window.
Sources: kunkafa.com · voltgridnews.com · amoledo.com · eastasiabrief.com
Data centers and energy
Single source: the IEA's 2026 report (during COP31) — capital spending by the five big tech companies topped $400B in 2025 and may rise another 75% to over $700B in 2026; global data-center electricity use was about 485 TWh in 2025 (+17%) and is projected at about 950 TWh by 2030; about 20% of planned data-center projects may be delayed by grid, transformer shortages and permitting (a single-source relay of the IEA report, with no link to the IEA original); Microsoft's 2026 environmental disclosure shows total greenhouse-gas emissions about 25% above the 2020 baseline, with electricity use up 24% in 2025; Amazon's purchased-power indirect emissions rose 34% in 2025. Single source: the FT reported on Nov 15 that Samsung SDS is in early talks with Anthropic and OpenAI to take them on as data-center tenants; analysts expect its data-center revenue to reach 6.6T won (about $4.9B) by 2031, about six times the current level (a single-source relay of the FT, whose original is paywalled; early-stage talks, no terms or capacity figures). Single source: LG Energy Solution and Samsung SDI announced on Nov 15 that they will show energy-storage and backup-power solutions for AI data centers at RE+ 2026 in Las Vegas on Nov 17–19; LG says it has been named a partner in NVIDIA's DSX Ready BESS program (company-announcement type news). The UK government launched a £180M AI data-center competition on Nov 11 (launched outside the window, reported inside it): to site the next national AI supercomputer, requiring proof of grid capacity, 10MW of IT load by Dec 31, 2027 and 30MW by March 2029; bids close at noon on Nov 25, with results expected Jan 8, 2027 (a single-source relay of Data Centre Review).
Sources: greentechlead.com · fourweekmba.com · koreawindintel.com · stechtimes.com
Capital markets
US and Hong Kong markets were closed on Sunday, Nov 15, with no new trading day; the Nov 13 closes were covered in the previous issue (S&P 500 7,561.67, -0.40%; Nasdaq 25,136.27, -0.91%; Dow 52,347.55, +0.10%; NVIDIA $162.85, -2.01%). No large new funding announcement in the window; the only small one is Tillermark announcing $1M from Sequoia and Techstars ($300K pre-seed plus $700K seed), single source. Single source: LangChain closed a $300M funding round, driven by enterprise agent-governance demand (unverified; Crunchbase's weekly top-10 list of Nov 13 also includes LangChain at $300M, which the previous issue did not itemize, so it is noted here). NVIDIA Q3 earnings are due after the close on Nov 17 (see NVIDIA tracking).
Sources: techfinancials.co.za · kimkj.com
Security incidents
Cursor's full disclosure (see Top Story 5). ARTEX detection signatures: independent researcher MalwareMustDie published an open-source ARTEX detection signature package on Friday, Nov 13 (reported by BleepingComputer); Nov 15 ARTEX coverage is follow-up re-reporting of the Nov 13 FBI FLASH alert, which was covered in the previous issue and is not repeated. No new security incident beyond ARTEX in the window.
Sources: aiworldwire.org · bleepingcomputer.com
Worth Reading
1."Meituan releases a 1.6-trillion-parameter agent model that never touched an Nvidia GPU" — The Decoder, 2026-11-15
How the LongCat-2.5 weights release came about, the 194-Atlas-A2 deployment details and the vendor-reported benchmark table — the verification source behind Top Story 3.
Sources: the-decoder.com
2."Megaspeed: How a little-known company became one of Asia's biggest buyers of Nvidia chips" — CNN, 2026-11-14
The three-year, over-$4B chip-buying investigation and the Alice Huang departure statement — the verification source behind Top Story 6.
Sources: cnn.com
3."17 attorneys general v. 13 BAAJA backers" — Hill Heat, 2026-11-15
The two cloture-vote times (about 18:00 on Nov 16 and about 01:00 on Nov 18) and the 17 state attorneys general's opposition letter — the verification source behind Top Story 7.
Sources: hillheat.news
Unverified Items
1.
Anthropic IPO refusing mainland China and Hong Kong investors: Bloomberg citing people familiar with the matter, via secondary reporting; the prospectus does not state the restriction in so many words, and the outcome of the BVI-trust application is undecided.
2.
Director Stevens' $946M stock sale: the Form 4 data is verifiable, but the reading of "no 10b5-1 plan" and the motive are media inferences; Buteon's Q3 total of about $968M differs slightly from the $946M figure.
3.
DeepSeek's first consumer paid plan ("Coding Plan"): only frontend code clues and community speculation; the company has confirmed no pricing, tiers or launch date. (Update: a Nov 14 check of the Nov 13 live build found zero hits for "Coding Plan", "plan", "subscription", "membership", "purchase" or "payment" across 603 JS chunks, and the official pricing page still lists only per-token API pricing; still no official answer, so it stays unverified.)
4.
Tencent planning to cut about 50% of Yuanbao and Hunyuan staff: resolved in this issue — Tencent PR director Zhang Jun said on Weibo on Nov 12 that it is "pure misunderstanding and a rumor" and "completely untrue"; the 50% figure was a "half army, half navy" metaphor from the Nov 11 anniversary livestream; see Top Story 9.
5.
Four Qwen4 versions with more than 3T parameters, an RSI self-recursion mechanism and Huawei Ascend support: claims from self-media, unconfirmed by Alibaba.
6.
Zhipu's final Nov 13 closing drop: only the intraday figure of -11.32% is seen, with the closing price unverified; the previous issue's -10.04% likewise had no source directly stating the close.
7.
Six Chinese models carrying Blackwell initialization "fingerprints": a one-sided research conclusion by former OpenAI engineer Will DePue, whose method has no independent verification and cannot prove where the chips actually are (they may have been used via US cloud providers).
8.
Revolut having about 2,000 engineers use GLM by default, cutting coding time by about 60% and costs by nearly 90%: only a single Chinese self-media repost, with no confirmation from Revolut or Zhipu.
9.
Microsoft 365 outage full-restoration timing: the previous issue said it was resolved on Nov 13, while a dedicated post-mortem puts full restoration in the early hours of Nov 14.
10.
GitHub Copilot retirement replacements: dev.to and Pondero say Fable 5 is replaced by Fable 5.2, while a Medium roundup writes Fable 5.1 — take GitHub's official changelog as authoritative; this piece did not open the official page to check.
11.
Nebius's up-to-$19.4B contract with Microsoft lagging in delivery in New Jersey: only one Nov 14 market analysis, with no other source cross-checking it.
12.
Edinburgh march against data centers: only Superpower Daily reported it; the 9.1 GW of added demand from 23 projects under review is a scenario estimate, not a forecast.
13.
Meta Muse profiling allegation and the Chrome extension: each single source, with no response or formal announcement from Meta.
14.
OpenAI's $15B round: CNBC citing sources, with the company not confirming it and the round not formally closed; the CFO acknowledged the pre-money valuation figure.
15.
Anthropic's RSP threshold determination: based on expectations about future models and the overall pace of progress, not on measurements of current models.
16.
Anthropic's nine-month revenue of $33.2B, operating loss of $1.3B and $70.7B in commitments due within 12 months: relayed from SEC filings via a weekly roundup, not checked line by line against the originals.
17.
The Megaspeed investigation being paused and the company never listed: a Nov 9 New York Times report via secondary reporting, with no new official developments in the window. (Update: CNN's Nov 14 investigation and Nov 15 report on Bessent blaming Lutnick are in Top Story 6; the "diversion to China" allegation remains "alleged" with no conviction, so it stays unverified.)
18.
BAAJA's two cloture votes (about 18:00 on Nov 16 and about 01:00 on Nov 18): resolved in this issue — Kunkafa's Nov 15 report gives the same schedule (first vote about 17:45–18:00 ET on Nov 16, second at about 01:00 on Nov 18); see Top Story 7.
19.
Anthropic IPO pricing Nov 18 and trading Nov 19: resolved in this issue — the Wall Street Journal confirmed on Nov 15 that pricing is Nov 18 and Nasdaq trading starts Nov 19 (ticker ANT), with a $70–78 price range; see Top Story 2.
20.
Anthropic IPO six-times oversubscription: resolved in this issue — Bloomberg, citing people familiar with the matter, says orders are about six times the shares offered, with CNBC and Fortune accounts agreeing; see Top Story 2.
21.
Grok 4.8 quiet rollout: only social posts and aggregator headlines claiming a limited rollout to Bot users on Nov 12, with no official confirmation from xAI (carried over).
22.
GPT-6.2 launch rumors: only X posts and screenshot rumors; OpenAI's official catalog has no such model (carried over).
23.
DeepSeek second-round financing figures: about 100B yuan (Caijing) versus 105.5B yuan (21st Century Business Herald and others), with pre-money valuation of about 500B yuan versus post-money 615.5B yuan, and no official confirmation (carried over).
24.
Unit 42 autonomous-intrusion report: a single-source chain (the Unit 42 report via CybersecurityNews and Reseller NZ), unverified (carried over).
25.
Zhipu ARR of $1.6B: no new corroboration found in the window, still single source and unverified (carried over).
26.
T-Head's confidential A1 filing: single source (China Fund News via Xueqiu), with the announcement text unverified (carried over).
27.
H3-Image license reading (not applicable in the US, EU, UK and South Korea; companies with annual revenue over $1M need separate written permission): a single-source reading, unverified (carried over).
28.
Epoch AI dataset (15.1 GW of leading data-center compute, 38 GW by end-2028): a single-source dataset, with power figures estimated, unverified (carried over).
29.
Chainstack AI-driven attack: single source, with the incident itself outside the window and the disclosure inside it, unverified (carried over).
30.
CoreWeave Q3 revenue figures: single-source aggregator reporting, with no Reuters or Bloomberg original link (figures fit the earnings-season timing, but unverified) (carried over).
31.
Claude Fable 5.2 fast mode on the Messages API: only one API aggregator briefing, unverified (carried over).
32.
Mirendil $1B Series A, UFORCE $700M Series A, Agile Robots $500M funding, Luminal $25M Series A and Databento $100M Series C: all single source or aggregator relays, unverified (carried over).
33.
Menlo Ventures report data: single source (21st Century Business Herald relaying a China Securities Journal reading of the Menlo Ventures report), with the original report data unverified (carried over).
34.
ByteDance AI chip chief Shi Yunfeng's resignation: single-source reporting, with no official response from ByteDance and the next destination unknown (carried over).
35.
Huang's Nov 18 closed-door briefing: no report of a public reply before the Nov 11 17:00 deadline; he resigned on Nov 12, and whether the briefing will take place or who will attend is unconfirmed; the discussion has now been superseded by the Senate Banking Committee testimony requirement of Dec 3 (carried over, updated).
36.
Whether Lutnick will be removed: Reuters citing sources that the White House is discussing it, with no official response; White House officials said on Nov 12 that Trump has "full confidence" in Lutnick, so the signals conflict and the matter is unconfirmed (carried over).
37.
Huang testifying Dec 3 and Lutnick Dec 10: the Senate Banking Committee summons is relayed by FourWeekMBA from NBC and CNBC, and Lutnick's agreement to appear is a single-source update; the House's Dec 1 closed-door interview required Huang to confirm in writing by 20:00 ET on Nov 13 or face a subpoena, with no report of the outcome in the window; Mast says Huang has agreed to testify before Thanksgiving on Nov 26, with the date still being negotiated (carried over). (Update: Lutnick has agreed to appear on Dec 10, per CBS via multiple relays; Huang's Dec 1 interview is covered in new item 60.)
38.
Whether NVIDIA approached Rene Haas: the FT report via Reuters, with Reuters saying it could not independently verify it; NVIDIA's board denies having formally contacted any candidate (carried over). (Update: the board said on Nov 13 that it is still selecting a search firm and has "not formally approached any candidate"; see NVIDIA tracking.)
39.
Anthropic IPO total raise figures: about $64.4B in the public offering (Reuters), with totals including the private placement given as "about $90.4B" (Yahoo Finance via Bloomberg) and "over $100B including the greenshoe" (FT) — take the official filing as authoritative (carried over).
40.
The seven co-founders selling about $1.75B of stock: Bloomberg via Pollar, saying each sold 3,378,378 shares for about $250M at the $74 midpoint, about $1.75B in total, with JonesTrading also marketing a block trade of up to $500M; single source, and the piece was published Nov 10 and updated Nov 13, so the timing boundary is unclear (carried over).
41.
Three senators' letters pressing Lutnick: only two secondary sources, with no authoritative reporting confirming the letters' contents (carried over).
42.
Senators Curtis and Blunt Rochester writing to Lutnick about a "remote access" loophole: AI Brief single source, unverified; OpenAI backing the Chip Security Act, MATCH Act and AI OVERWATCH Act: a Nov 12 Bloomberg report via WinBuzzer, a single-source chain (carried over).
43.
Bloomberg's "restricted NVIDIA chips traced to a Shanghai data center": Engadget and Seeking Alpha relaying Bloomberg; the documents do not prove Alice Huang herself took part in chip diversion; Megaspeed has bought more than $4.8B of NVIDIA chips since June 2024, unverified independently (carried over).
44.
LITONG moving 240,000 NVIDIA chips (at least $7.5B) into China: only in a relay of The Information's reporting, with no other independent source, and the claim involves smuggling allegations, so it is marked unverified (carried over).
45.
DeepSeek-linked investors under US investigation: a KuCoin flash report saying companies linked to Alice Huang invested in DeepSeek around Nov 12 and promised NVIDIA chip access, while her Singapore-based Megaspeed International is under US investigation over NVIDIA chip imports worth billions of dollars, with NVIDIA's board also reviewing the matter; the piece itself says the intelligence-link claim "remains only a question". Only a single aggregator flash, with no mainstream corroboration, and the facts and figures are unverified — not suitable as main-text news (carried over).
46.
NVIDIA's Nov 13 closing price figures: Yahoo Finance and Finnhub give $162.85 (-2.01%), while individual quote sources report $161.24, $161.52 or $161.77, likely after-hours or data-vendor differences; the main text uses $162.85 (carried over, updated).
47.
NVDA's Nov 12 closing price figures: ChartExchange recorded $164.27 (-5.45%), versus $166.19 (-4.34%) from Finnhub and several closing roundups — Finnhub is used here (carried over).
48.
TSMC's two investments of up to $500M each to take part in Anthropic's IPO: single source (Korben.info), unverified.
49.
OpenAI disbanding its independent robotics team for the second time: single source (a daily digest), with no official announcement or mainstream corroboration.
50.
Claude Meetings beta: single source, unverified.
51.
Claude Opus 5.5 as the default model for all paid plans: single source, unverified.
52.
Gemini 4 Carbon topping the MathArena benchmark: single source, unverified.
53.
Muse accounting for 59% of agent requests seen by HUMAN Security: a single vendor report, unverified.
54.
Meta internally testing a Llama-powered prediction-market app: single source, unverified.
55.
Muse Ember inference model and Muse Wonder: single-source LinkedIn newsletter, with the publication date not stated, unverified.
56.
A second lawsuit seeking to freeze AI development: single low-quality aggregator, unverified.
57.
LangChain's $300M funding round: single source, unverified.
58.
Manus paid-promotion controversy: single-source mirror of a Verge report, unverified.
59.
Jensen Huang "considering a return" as NVIDIA CEO: FT original via multiple relays, but NVIDIA and Huang declined to comment and there is no board announcement; petition signature counts differ across sources (over 100,000 / 115,000 / over 156,000).
60.
Huang's Dec 1 House interview: oral "agreed" claims by chairs Mast and Moolenaar conflict with no public written confirmation as of the evening of Nov 14; correction — the written-confirmation deadline was 20:00 ET on Nov 13, not Nov 14 as the previous issue's main text stated.
61.
BAAJA procedural vote counts: 73–25 versus 62–21, possibly referring to different motions (motion to proceed versus cloture on the amended text).
62.
The Megaspeed diversion allegation remains "alleged" with no conviction; Bessent blaming Lutnick for delaying the Entity List listing is a single-source claim from anonymous officials.
63.
IEA data-center report figures: single-source relay, with no link to the IEA original.
64.
Samsung SDS in early talks with Anthropic and OpenAI as data-center tenants: single-source relay of the FT, early stage with no terms or capacity figures.