Shared research report

What are the most significant developments in AI today?

November 15, 2026

Research Report

Question: What are the most significant developments in AI today?

Date: 2026-11-15T20:20:50.733278690+00:00

Coverage window: 2026-11-15 – 2026-11-15

Rounds: 4

Status: PARTIAL

Objective check — 1 of 4 criteria met

The run produced work, but the objective below is not fully achieved. Each unmet criterion names what is still outstanding.

Evidence: 46 claims · 39 sourced · 0 partial · 0 unsupported · 7 self-reported (no independent source) · 7 single-source

Executive Summary

As of 2026-11-15, the most significant AI development is a policy/market intervention, not a model release: U.S. Treasury Secretary Scott Bessent publicly called for Nvidia to hire an outside CEO and demanded that the board and outside counsel answer within 10 days, in a Fox Business interview published at 2:25 p.m. EST on 15 November 2026 (Fox Business; same-day Reuters headline: Bessent says Nvidia must “go outside the company” to find new CEO).

The underlying Nvidia crisis — Jensen Huang’s resignation, Colette Kress as interim CEO, Harvey C. Jones as board chair — is dated 12 November, so today’s news is external pressure applied to a three-day-old event (Nvidia press release, 12 Nov; 8-K accepted 2026-11-12 08:52:41).

No verified model, product, benchmark, or peer-reviewed research release is dated 2026-11-15. arXiv’s newest cs.AI submission at retrieval was 2026-11-12T18:59:06Z and a 15 November date query returned zero; Google’s sitewide RSS newest item was 13 Nov 20:30 UTC; DeepMind’s newest was 12 Nov 18:00:50 UTC; Hugging Face Daily Papers’ newest edition was 13 Nov; Nvidia’s own newsroom was current to 13 Nov (arXiv API; Google RSS; DeepMind RSS; HF Daily Papers; Nvidia newsroom).

What Is Actually Dated 2026-11-15

DevelopmentWhat happened on 2026-11-15EvidenceStatus
Bessent intervenes in Nvidia CEO searchTreasury Secretary Scott Bessent said Nvidia’s board should “go outside the company” and needs “a dramatic shake-up”; demanded answers “in the next 10 days from the board and from the outside counsel.”Fox Business datePublished: 2026-11-15T14:25:00-05:00; Reuters same-day story slugPublished statement verified. No Treasury primary release located; the 10-day clock’s start date is not established.
Megaspeed statement on Alice HuangMegaspeed said Alice Huang left the company in August 2025 — a 15 November disclosure about a 15-month-old departure.Reuters, 2026-11-15Headline/index verified; body not read. Not a fresh executive change.
Korea tax-exam AI cheating caseThe top scorer on South Korea’s tax accountant exam confessed to using AI to cheat and turned himself in to police.Reuters, 2026-11-15Headline verified. Enforcement/disciplinary matter, not AI policy.
SEBI signals AI guidelines for capital marketsSEBI Chairperson Tuhin Kanta Pandey reportedly said the regulator “will soon issue detailed guidelines” on responsible AI use in capital markets.Secondary aggregation dated 15 Nov, citing prokerala.com: theworldofai.orgLow confidence. No SEBI press release or speech transcript verified; status of the guideline is unverified.
Research/product releasesNone verified. arXiv date-bounded query returned totalResults = 0; Google/DeepMind RSS had no 15 Nov item; HF Daily Papers had no 15 Nov page.arXiv date query; Google RSS; DeepMind RSSVerified negative across those surfaces. Not proof nothing shipped industry-wide — OpenAI/Anthropic/Meta/xAI/Mistral/Cohere were not exhaustively checked.

The Recirculation Trap: What Looks Like Today’s News but Isn’t

Most “November 15, 2026 AI news” is repackaged 12–13 November material. Date each item to the instrument, not the coverage peak.

Story presented as 15 NovActual datePrimary evidence
Gemini 4 Hydrogen launch12 Nov 2026, 18:00 UTC; modified 13 NovGoogle blog JSON-LD: datePublished: 2026-11-12T18:00:00+00:00; Wayback first capture 2026-11-12 18:03:05 UTC (Google blog)
Nvidia CEO resignation / Kress interim CEO / Jones chair12 Nov 2026; 8-K accepted 12 Nov 08:52:41, period of report 11 NovNvidia press release; 8-K index
OpenAI ChatGPT Wallet rollout~13 Nov 2026 — TechCrunch dated it two days before 15 NovTechCrunch AI section
Philippines AI governance executive orderSigned 12 Nov 2026 as Executive Order No. 131; 15 Nov coverage was the peakSecondary aggregation: theworldofai.org — gazette text unretrieved
Nonprofit lawsuit to halt frontier AI (Irreplaceable v. Anthropic et al.)Filed 12 Nov 2026 in San Francisco Superior CourtAs-filed PDF last-modified: 2026-11-12 14:52:20 GMT; Public Citizen release 2026-11-12T16:39:07Z (PDF)
“AI Papers, November 15, 2026” digestPapers submitted 12 Nov 2026arXiv 2611.16591 and 2611.17108 submission histories (2611.16591; 2611.17108)

Analysis

The day’s shape is explained by the calendar: 2026-11-15 was a Sunday. arXiv does not announce Friday or Saturday, EDGAR does not accept filings on weekends, and labs generally do not launch models on Sundays. The result is a news trough filled by reaction coverage of a Thursday/Friday cluster: Nvidia’s 12 November leadership change, Gemini 4 Hydrogen’s 12 November launch, the 12 November frontier-AI lawsuit, and the 12 November Philippines executive order.

The Nvidia thread is the week’s dominant AI story, but its center of gravity is 12 November, not today. Nvidia’s own release says Huang resigned “effective immediately,” Kress became interim CEO while remaining CFO, and Harvey C. Jones became the board’s first-ever chair (Nvidia, 12 Nov). The 15 November development is Bessent’s televised demand for an outside CEO and a 10-day answer — a political escalation, not a new corporate act.

For research, the correct statement is narrower than “nothing happened”: no 2026-11-15-dated paper had been announced or indexed by roughly 20:10 UTC on 15 November, and arXiv’s schedule means a 15 November submission would not have appeared in a 15 November announcement (arXiv availability policy).

Risks & Open Questions

Claims without independent support

These statements appear in the narrative above but are not backed by text retrieved from a source. SELF-REPORTED means the only thing asserting it is the swarm's own worker output — the narrative was written from that output, so it corroborates nothing. Treat all of these as unverified.

Bibliography

  1. Fox Business
  2. Bessent says Nvidia must “go outside the company” to find new CEO
  3. Nvidia press release, 12 Nov
  4. 8-K accepted 2026-11-12 08:52:41
  5. arXiv API
  6. Google RSS
  7. DeepMind RSS
  8. HF Daily Papers
  9. Nvidia newsroom
  10. Reuters, 2026-11-15
  11. Reuters, 2026-11-15
  12. theworldofai.org
  13. arXiv date query
  14. Google blog
  15. TechCrunch AI section
  16. PDF
  17. 2611.16591
  18. 2611.17108
  19. arXiv availability policy
  20. Anthropic news

Detailed Findings

Round 0 · Finding 1

AI Research Published 2026-11-15 — In-Window Verification Report

Research conducted: 2026-11-15, ~19:20–20:10 UTC (timestamps returned by the arXiv API during this session). Bottom line up front: I could not verify three (or even one) research result published on 2026-11-15 from a primary source. The sources that look like 2026-11-15 research drops are aggregator digests whose date is the digest date, not the publication date — and the papers behind them were submitted 12 Nov 2026. This report documents that finding rather than filling the gap.


1. Executive Summary

Therefore: the mission's success criterion of "at least three distinct AI developments explicitly dated 2026-11-15" is not met, and I am not going to manufacture it. What follows is the verified negative/structural finding plus properly labelled background.


2. Key Findings

Finding A — The "2026-11-15" arXiv research digest is a repackaging of 12 Nov submissions (Confidence: HIGH)

A page titled "AI Papers, November 15, 2026" with the byline "November 15, 2026 · 3 papers" lists three arXiv papers (thinksmart.life/papers/2026-11-15/). I fetched each arXiv abstract page to check the real publication date:

…(truncated — the summary above captures the substance)

Round 0 · Finding 2

Executive Summary

In-window, verifiable AI policy/legal developments dated 2026-11-15 are thin — and I could not verify any of them against a primary government or court record. The single most important structural finding is calendrical: 2026-11-15 was a Sunday, which suppresses official press releases, court rulings, and filings. Two fetched pages independently confirm the weekday: theworldofai.org's news index is stamped "last compiled Sun, 15 Nov 2026 18:25 UTC" (https://theworldofai.org/ai-news/), and AIBriefs' page title is "AI News Today: Sunday, November 15, 2026" (https://aibriefs.news/ai-news-today, bot-blocked on fetch; weekday taken from the search-result title, not from a page I read).

Critically, the majority of what surfaced as "November 15, 2026 AI news" is press recirculation of 12–13 November announcements — precisely the failure mode the mission warned about. I document those explicitly as excluded rather than silently dropping them, and I flag one that a naive reading would have counted as a policy win.

I found one policy/regulatory item that appears genuinely to originate on 2026-11-15, and zero court rulings, standards-body decisions, or lab safety publications dated that day. I state that as a gap, not as a null result I inferred.


Key Findings

Finding 1 — POLICY/REGULATION: India's SEBI signals imminent AI guidelines for capital markets (reported 2026-11-15)

Confidence: LOW–MODERATE. Dated reporting exists; primary record unverified.

SEBI Chairperson Tuhin Kanta Pandey said the regulator "will soon issue detailed guidelines on the responsible use of Artificial Intelligence in capital markets," per reporting dated November 15, 2026 (https://theworldofai.org/ai-news/, item uid 602bae5f, tagged Policy & Regulation, carried by 8 outlets, attributed to prokerala.com on November 15, 2026).

Finding 2 — POLICY/REGULATION: Philippines AI governance executive order — EXCLUDED, pre-dated

The most widely carried "Nov 15" policy story fails the constraint.

The top-ranked story on the 15 November index was "President Marcos signs executive order setting direction for AI governance in PH" (10 outlets, tagged 2026-11-15) — but the same entry states: "He signed Executive Order No. 131 on November 12, 2026" (https://theworldofai.org/ai-news/, uid bbba87d8).

…(truncated — the summary above captures the substance)

Round 0 · Finding 3

AI Business Developments — 2026-11-15 (in-window only)

Executive Summary

The in-window business news flow for 2026-11-15 was exceptionally thin, and I could not verify a single funding round, acquisition, partnership, earnings disclosure, or major customer deal announced or first reported that day. This is a dated, checkable negative rather than an absence of searching: 2026-11-15 was a Sunday (confirmed by AI Weekly's own page header, "Sunday, November 15, 2026" — https://aiweekly.co/ai-news-today), and the two outlets I could audit at the article level published almost nothing.

The concrete evidence:

What is verifiable on 2026-11-15 is a governance/policy shockwave around Nvidia, plus a data-integrity problem in the AI funding-tracker ecosystem that makes most circulating "November 2026 AI funding" numbers unusable.


Key Findings

1. Business / governance — US Treasury Secretary calls for an outside Nvidia CEO (VERIFIED, dated 2026-11-15)

Reuters published on 2026-11-15 a story headlined "Bessent says Nvidia must 'go outside the company' to find new CEO," with the summary: "The company needs 'a dramatic shakeup,' the treasury secretary told Fox News." This is a US cabinet official publicly intervening in the CEO succession of the world's most important AI chip company — a business-governance event with direct market relevance.

Confidence: high that Reuters published this on 2026-11-15 (headline, summary and 2026-11-15 URL slug all visible on the fetched index). Caveat: I verified the headline and dek from Reuters' own index page, not the article body.

…(truncated — the summary above captures the substance)

Round 0 · Finding 4

AI Developments Dated 2026-11-15 — Verification Report

Bottom line: I could not verify a single major model or product release dated 2026-11-15 from any primary lab source. The date fell on a Sunday, and every primary source I could reach either showed no 11-15 publication or was inaccessible. I am reporting that negative result rather than filling the gap, because the available "November 2026 release tracker" pages contradict each other on model names and are not trustworthy.


1. Executive Summary

Mission success criterion NOT met. The brief required at least three distinct developments explicitly dated 2026-11-15. I found zero that I can source to a dated primary record. I am flagging the shortfall rather than inventing three.


2. Key Findings

Finding A — Product/model releases: NO verified release on 2026-11-15

Confidence: HIGH for Anthropic; UNVERIFIED for everyone else.

Anthropic's announcements index (79 articles) lists, in reverse chronological order: Nov 10 (IPO roadshow), Nov 9 (Claude Fable 5.2 + Claude Mythos 5.2), Oct 28, Oct 26, Oct 22, Oct 21, Oct 18, Oct 14, Oct 7 (Claude Haiku 5.5), Oct 6, Oct 2. There is no Nov 15 entry. Source: https://www.anthropic.com/news/announcements (fetched 2026-11-15).

…(truncated — the summary above captures the substance)

Round 1 · Finding 1

arXiv cs.AI new batch and Hugging Face Daily Papers for 2026-11-15 — plus the MAPE paper's true submittedDate

Retrieval window: all primary fetches below were executed between 2026-11-15T20:10:58Z and 2026-11-15T20:12Z, i.e. inside the target window. HTTP response headers from the arXiv API independently confirm the clock (date: Sun, 15 Nov 2026 20:10:58 GMT).


Executive Summary

Three findings, in order of importance:

  1. arXiv 2611.17068 (MAPE) self-dates to 12 November 2026 — neither the 15th nor the 16th. The abs page's own submission history reads [v1] Thu, 12 Nov 2026 16:42:04 UTC. The question as posed in the open gap (15th or 16th) is a false dichotomy; both candidate answers are wrong. The paper is out-of-window context, not an in-window finding.

  2. There is no 2026-11-15 arXiv cs.AI batch, and no 2026-11-15-dated cs.AI paper is publicly visible at all. arXiv's own announcement schedule makes this a structural certainty, not an accident of timing: Sunday-dated submissions are announced Monday 20:00 ET = 01:00 UTC Tuesday 17 November — roughly 29 hours after this retrieval. The premise that a Sunday batch would carry 15-November items is contradicted by the issuing body's own published schedule.

  3. Hugging Face has no page for 2026-11-15. huggingface.co/papers/date/2026-11-15 redirects (HTTP 200 after redirect) to /papers/date/2026-11-13. /papers/date/2026-11-16 redirects to the identical target. Verified absence, with the failure mode named.

The honest bottom line: this sub-question yields a dated negative, not a development. Nothing on these two surfaces is a 2026-11-15 publication.


Key Findings

Finding 1 — MAPE (arXiv 2611.17068) is dated 12 November 2026. Confidence: very high

The abs page at https://arxiv.org/abs/2611.17068 carries, verbatim:

Title: "Discovering Effective Agent Memory Designs through Program Evolution" (MAPE). Authors: Xinheng Ding, Seyyed Saeid Cheshmi, Ziyang Zhang, Tairan Chen, Wenya Xie, Yu Hou, Rui Zhang, Zirui Liu. Comments: 24 pages, 8 figures, 10 tables.

Classification: OUT-OF-WINDOW CONTEXT (12 Nov 2026 falls outside 2026-11-15..2026-11-15). It must not be reported as a 15 November development.

Two structural details worth recording, because they explain why this paper surfaced in a digest but never appeared in the cs.AI new list:

…(truncated — the summary above captures the substance)

Round 1 · Finding 2

Nvidia's primary record for 2026-11-15 — findings

1. Executive summary

Nvidia's own primary record contains nothing dated 2026-11-15. The company's newsroom was live and current through November 13 but carries no November 14 or November 15 entry, and EDGAR full-text search returns exactly zero 8-K filings mentioning NVIDIA anywhere in the 2026-11-14 → 2026-11-16 range. The leadership event itself is fully documented, and it is dated 2026-11-12 (press release, GlobeNewswire dateline) with an 8-K accepted 2026-11-12 at 08:52:41 and a period of report of 2026-11-11.

The only thing that verifiably happened in AI-corporate terms on 2026-11-15 in this thread is not a corporate act at all: Treasury Secretary Scott Bessent's televised call for an outside CEO, published by Fox Business at 2026-11-15 14:25 EST. The "congressional-testimony agreement" could not be located in any primary record and must be treated as unverified, not as an 11-15 event.

The single most important correction this round forces: the Nvidia story was already four days old on 2026-11-15. The 11-15 item is a Treasury comment about a 11-12 ouster, plus a 10-day demand for answers — a forward-looking deadline, not a same-day corporate development.


2. Nvidia-owned surfaces: verified, live, and silent on the 15th

Both newsroom surfaces were fetched successfully and render current content — so the negative is a real negative, not a stale mirror or a broken feed.

https://nvidianews.nvidia.com/news — HTTP 200. The most recent entry is dated November 13, 2026 ("How AI Is Helping Care Teams Get Patients the Right Care, Sooner", a blog post). The next entries are November 12, 2026 (GeForce NOW blog; and the press release "NVIDIA Announces CEO Transition"), then November 10, 2026 (the independent-directors' statement re: Wachtell, Lipton, Rosen & Katz), then November 8, 2026 (board statement; "A Message to Our Company"). No item is dated November 14 or November 15.

https://nvidianews.nvidia.com/news/latest — HTTP 200, identical ordering, same top date of November 13, 2026, and the same absence of any November 14/15 entry.

This is the key structural point: the index was actively publishing through the 13th. A feed that has stopped updating and a feed with nothing to say look identical in a screenshot but mean opposite things; here the first is excluded, so the silence on the 14th–15th is meaningful.

Confidence: high. Two independent Nvidia-owned surfaces, both HTTP 200, both current to Nov 13.


3. The CEO transition: a 2026-11-12 event, with an 11-11 period of report

Press release — "NVIDIA Announces CEO Transition", dated November 12, 2026. Fetched at https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-CEO-Transition/ (HTTP 200), and indexed on the newsroom at https://nvidianews.nvidia.com/news/nvidia-announces-ceo-transition.

…(truncated — the summary above captures the substance)

Round 1 · Finding 3

Machine-readable surface sweep: six labs, 2026-11-15

Scope note up front: this round attacked only the six lab-owned machine surfaces named in the task. Four of the mission-manifest success criteria (an Nvidia-owned primary artifact, arXiv 2611.17068's submittedDate, the SEBI / Philippines EO 131 / BIS policy instruments, and a non-Western-lab surface) were not attempted in this round and are reported at the end as open, unverified gaps — not as findings, and not sorted into any category.

Executive Summary

No lab among OpenAI, Google/Google DeepMind, Meta AI, xAI, Mistral or Cohere published an item dated 2026-11-15 on any machine-readable surface I could reach. The strongest result is a live, in-window dated negative for OpenAI: its news feed was regenerated on 15 Nov 2026 at 19:51:07 GMT and still carried nothing newer than 13 Nov. Meta AI yielded no machine surface at all (two 404s), which is a verified absence of the surface, not evidence of silence. Cohere and xAI returned 200s whose bodies were truncated by the retrieval layer, so their negatives are partial rather than exhaustive. In-window lab-announcement evidence is therefore thin — and I am not filling that gap from memory.

Key Findings

#FindingConfidence
1OpenAI published nothing dated 2026-11-15 on openai.com/news/rss.xml; feed rebuilt 15 Nov 19:51:07 GMT, newest item 13 NovHigh
2Google's AI-section feed carried nothing after 12 Nov 18:00 GMT; feed itself was not rebuilt on the 15thMedium-High (AI section only)
3Meta AI exposes no reachable sitemap or blog RSS — both 404High (as a surface negative)
4xAI's sitemap lists no news item after 2026-11-02Medium-High
5Mistral's /news/rss.xml is 404; its sitemap index is frozen at 2026-11-12T10:01:57.692ZMedium (index not fetched)
6Cohere's sitemap shows no 2026-11-15 lastmod in the retrieved portionLow-Medium (truncated)
7Wayback CDX for the exact openai.com root on 20261115 returned an empty setMedium (narrow query scope)

Detailed Analysis

1. OpenAI — 200, live on the day, nothing new (dated negative)

https://openai.com/news/rss.xml → HTTP 200, served over plain HTTP.

The channel-level build timestamp is Sun, 15 Nov 2026 19:51:07 GMT — an in-window timestamp proving the feed was regenerated on the 15th. The newest item beneath it is:

followed by https://openai.com/business/customer-stories/shield (13 Nov 08:00 GMT), .../amplitude (12 Nov 20:00 GMT), .../gondola (12 Nov 12:00 GMT), https://openai.com/index/children-first-declaration (12 Nov 00:00 GMT), https://openai.com/index/chatgpt-plus-for-veterans (11 Nov 08:00 GMT), and so on back to late October.

…(truncated — the summary above captures the substance)

Round 1 · Finding 4

Policy primary-record check for 2026-11-15

Bottom line

Across the three policy threads, none of the issuing bodies published anything verifiable on 2026-11-15. Two of the three bodies could not be read at all (bot-wall / 404), and the one machine-readable primary record I could query — the Federal Register API — returned a hard zero for that date. The most important correction this round produces is negative: Philippines EO No. 131 is a 2026-11-12 instrument, and it remains unverified at the issuing body; it must not be reported as a 15 November development.


1. SEBI — chair Tuhin Kanta Pandey's AI guidelines for capital markets

Result: NOT FOUND on 2026-11-15. No press release, speech or circular page on sebi.gov.in carrying the AI guidelines.

Surfaces tried:

SurfaceURLStatusOutcome
SEBI Circulars listinghttps://www.sebi.gov.in/sebiweb/home/HomeAction.do?doListing=yes&sid=1&ssid=7&smid=0HTTP 200Page loaded, but extraction returned only site navigation chrome; no dated circular rows retrievable, no AI circular visible
SEBI Filings listing (wrong section, checked anyway)https://www.sebi.gov.in/sebiweb/home/HomeAction.do?doListing=yes&sid=3&ssid=6&smid=0HTTP 200Navigation chrome only; not a press-release surface
SEBI homepage tickerhttps://www.sebi.gov.in/ (via Bing organic result, domain www.sebi.gov.in)HTTP 200Ticker items captured reference World Investor Week 5–11 Oct 2026 and a Pandey message for that occasion — no AI-guidelines item

The site-scoped search for Pandey's speeches (site:sebi.gov.in artificial intelligence guidelines Pandey speech, DuckDuckGo) returned only Pandey speeches from Feb, Mar, Aug 17, Sep (Global Fintech Fest 2026) and Oct 2026 (SAMVAAD India Open Finance Summit 2026) — e.g. https://www.sebi.gov.in/media-and-notifications/speeches/oct-2026/address-by-shri-tuhin-kanta-pandey-chairman-sebi-samvaad-india-open-finance-summit-2026-sahamati-annual-conference_105276.html. No November 2026 speech and no AI-guidelines-for-capital-markets speech appeared in the index.

Caveat, stated plainly: I did not paginate SEBI's full speeches archive, and absence from a search index is weak evidence. So the honest finding is: the "soon" framing has no corroborating SEBI-published artifact that I could locate on 15 November, and the guidelines remain speech-derived. This is a verified absence of a findable document, not proof that no internal drafting exists.

One lead worth chasing, undated: SEBI hosts an AI document on its own domain — "Frontier AI readiness: Defining the next era of enterprise resilience" at https://cybersuraksha-ai.sebi.gov.in/documents/48365173/0/Frontier_AI_Readiness_Enterprise_Resilience.pdf/... (surfaced in search; not fetched, no date established). Do not treat this as a 15 November artifact.


2. Philippines Executive Order No. 131 — signing date and text

…(truncated — the summary above captures the substance)

Round 2 · Finding 1

Nvidia Leadership Upheaval, 2026-11-12 → 2026-11-15: What Primary Records Actually Exist

Executive Summary

There is a genuine primary record for the leadership change itself: an NVIDIA newsroom / GlobeNewswire press release dated November 12, 2026, which confirms both names in the question — Colette Kress as Interim CEO and Harvey C. Jones as the first-ever Chair of the Board. That claim is VERIFIED.

Almost everything else in the question is not verified by any primary record I could retrieve. I found no SEC filing, no Treasury press release substantiating a "10 days" demand, and no committee announcement, hearing notice, or company statement confirming a congressional-testimony agreement. A key structural finding: NVIDIA's own newsroom index shows the governance crisis was already public before the window — a board statement and an all-company message on November 8, and the independent directors' statement naming Wachtell Lipton on November 10. The 12 November release is the resolution of a story that broke earlier, not its origin.

Important honesty note on the freshness rule: you defined in-window as 2026-11-15 only. I retrieved no primary record dated 2026-11-15. Every primary document below is dated 2026-11-12 or earlier, i.e. inside the mission's 11-12→11-15 window but outside the strict single-day window. I am labelling them accordingly rather than re-dating them.


Key Findings

1. VERIFIED — Interim CEO and board chair, from the company's own release (2026-11-12)

The NVIDIA newsroom release "NVIDIA Announces CEO Transition," dated November 12, 2026 (dateline "SANTA CLARA, Calif., Nov. 12, 2026 (GLOBE NEWSWIRE)"), states verbatim:

"NVIDIA (NASDAQ: NVDA) Board of Directors announced today that Jensen Huang has resigned as President, Chief Executive Officer, and a member of the Board, effective immediately. Colette Kress, current Executive Vice President and Chief Financial Officer of NVIDIA, has agreed to serve as Interim Chief Executive Officer. Kress will continue to serve as Executive Vice President and Chief Financial Officer while serving as Interim Chief Executive Officer until a permanent successor has been named."

And on the chair:

"The Board of Directors appointed Harvey C. Jones as its first-ever Chair of the Board. Mr. Jones has been an entrepreneur, technology executive, and active venture investor for over 30 years, including serving as Chief Executive Officer and Executive Chairman of Synopsys."

Source (body text retrieved in full, with visible date): https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-CEO-Transition/default.aspx Canonical newsroom URL: https://nvidianews.nvidia.com/news/nvidia-announces-ceo-transition

…(truncated — the summary above captures the substance)

Round 2 · Finding 2

Lawsuit to Halt Frontier AI Development — What the Primary Record Actually Shows

Sub-question scope: the nonprofit suit filed on or about 2026-11-12 in San Francisco. Research cutoff: 2026-11-15.


Executive Summary

Four nonprofit organizations — Irreplaceable Education Fund, Public Citizen, Heat Initiative, and The Alliance for Secure AI — filed suit in the Superior Court of California, County of San Francisco on Thursday, 2026-11-12, against Anthropic, Google, Meta, OpenAI (both the OpenAI Group and the OpenAI Foundation), and two "SpaceX Super Intelligence" (SpaceXSI) entities. The caption is Irreplaceable et al. v. Anthropic et al. I retrieved the filed pleading itself as a PDF (63 pages, hosted by plaintiffs' counsel Platkin LLP, last-modified: Thu, 12 Nov 2026 14:52:20 GMT) and Public Citizen's own announcement (datePublished 2026-11-12T16:39:07+00:00).

Critically, I could not extract the pleading's text. The PDF is served as compressed binary streams and my fetch returned raw PDF object data, not rendered text. Therefore the case number, the enumerated causes of action, the standing theories, the specific relief prayed for, and whether a TRO or preliminary injunction was formally sought are NOT verified by me from the complaint text. I report below only what is established by a primary artifact, and I flag every element that rests on secondary reporting.

One structural correction to the manifest's assumption: because this is a California state trial court filing, it would not appear on PACER (federal courts only), and CourtListener/RECAP — which mirrors federal PACER — would not carry it either. The correct primary channels are the San Francisco Superior Court's own case-information portal and plaintiffs' counsel's filed copy. That is why the docket-specific details the mission asked for are hard to obtain here.


Key Findings

1. The filed pleading exists and is retrievable — verified

…(truncated — the summary above captures the substance)

Round 2 · Finding 3

Two Reported Policy Actions: Verification Status as of 2026-11-15

Executive Summary

Neither of the two reported policy actions can be confirmed from a primary record as of 2026-11-15.

(a) The Commerce/BIS "AI remote-access rule" is NOT VERIFIED as finalized. The Federal Register — queried directly through its public API — contains no such document. Every BIS document published between 2026-10-20 and 2026-11-15 is enumerable and none concerns remote access to AI compute. The underlying reporting itself uses the word "finalizing," which describes an anticipated action, not a completed one. This appears to be a genuine gap between secondary reporting and the primary record.

(b) Philippines EO No. 131 is reported consistently by multiple outlets as signed 2026-11-12, three days before the 2026-11-15 Philstar story — but the Official Gazette text could not be retrieved. The gazette domain is bot-blocked, and the Internet Archive (the fallback the mission suggested) returned HTTP 503 "Temporarily Offline" during this session. The signing date and the Council mandate below therefore rest on secondary reporting only and must be labeled as such.


Key Findings

(a) BIS AI remote-access rule — UNVERIFIED / no primary record exists

CheckResultConfidence
Federal Register, all BIS docs 2026-10-20→2026-11-154 documents, none AI remote-accessHigh
Federal Register, all agencies, full-text "remote access" 2026-10-15→2026-11-157 documents, none AI-relatedHigh
Federal Register, BIS + "remote access" artificial intelligence (all time)4 docs, newest 2026-01-15High
reginfo.gov RIN/OMB docket lookupNOT COMPLETED — could not query in this sessionn/a
Wayback capture of bis.gov press releasesNOT COMPLETED — Internet Archive returned 503n/a

The complete list of BIS documents published in the 2026-10-20 → 2026-11-15 window, from the Federal Register API (https://www.federalregister.gov/api/v1/documents.json?conditions[agencies][]=industry-and-security-bureau&conditions[publication_date][gte]=2026-10-20&conditions[publication_date][lte]=2026-11-15):

  1. 2026-22793 — "Extension of the Suspension of Expansion of End-User Controls for Affiliates of Certain Listed Entities," Rule, 2026-11-09, Docket No. 261027-0002 — https://www.federalregister.gov/documents/2026/11/09/2026-22793/extension-of-the-suspension-of-expansion-of-end-user-controls-for-affiliates-of-certain-listed
  2. 2026-22726 — Paperwork Reduction Act notice on polysilicon import qualification, Notice, 2026-11-06
  3. 2026-22725 — Paperwork Reduction Act notice on Section 232 specialty pharmaceuticals, Notice, 2026-11-06
  4. 2026-21840 — "Order Renewing Order Temporarily Denying Export Privileges," Notice, 2026-10-27

…(truncated — the summary above captures the substance)

Round 2 · Finding 4

Findings: Release-surface sweep, 2026-11-10 → 2026-11-16

Method note. I swept Hugging Face org pages for all six named non-Western labs and, because the org-page "Updated N days ago" label proved to be a last-modified signal rather than a release date, I re-ran each org through the Hugging Face models API sorted by createdAt descending — that field is the repo creation timestamp and is the only reliable release date on the surface. All API calls were made 2026-11-15 20:16 UTC.


1. Non-Western labs: no model release in the window (VERIFIED NEGATIVE)

Across all six orgs, the newest createdAt on 2026-11-15 falls outside 2026-11-10..11-16. There is no in-window release from any of them.

LabNewest repocreatedAt (UTC)In window?
DeepSeekdeepseek-ai/DeepSeek-V4.1-Flash-10172026-10-17T02:40:38ZNo (−24d)
Alibaba/QwenQwen/Qwen-Image-2.1-Turbo2026-10-09T04:50:01ZNo (−32d)
Moonshotmoonshotai/Kimi-K3.1-Pro2026-11-02T04:13:51ZNo (−8d)
Zhipu / Z.aizai-org/GLM-5.3-Flash-BF162026-08-25T06:45:05ZNo (−77d)
Baidubaidu/Unlimited-OCR-22026-10-26T03:57:46ZNo (−15d)
MiniMaxMiniMaxAI/MiniMax-H3-Image-PE2026-11-08T15:35:23ZNo (−2d)

Sources (each fetched directly): https://huggingface.co/api/models?author=deepseek-ai&sort=createdAt&direction=-1&limit=8 · https://huggingface.co/api/models?author=Qwen&sort=createdAt&direction=-1&limit=8 · https://huggingface.co/api/models?author=moonshotai&sort=createdAt&direction=-1&limit=8 · https://huggingface.co/api/models?author=zai-org&sort=createdAt&direction=-1&limit=8 · https://huggingface.co/api/models?author=baidu&sort=createdAt&direction=-1&limit=6 · https://huggingface.co/api/models?author=MiniMaxAI&sort=createdAt&direction=-1&limit=6

The trap this sweep avoided — worth flagging explicitly. The rendered org pages give misleading recency signals that would have generated three false in-window "releases":

So: none of the six labs shipped a model in the 2026-11-10→11-16 window as of 2026-11-15. The manifest's non-Western blind spot resolves to a genuine lull, not a missed launch. Highest confidence finding in this report.


…(truncated — the summary above captures the substance)

Round 3 · Finding 1

Findings: publication date of "Gemini 4 Hydrogen" and Google/DeepMind output dated 2026-11-15

1. The primary permalink settles it: Gemini 4 Hydrogen is dated 2026-11-12, not 2026-11-15

I fetched the announcement permalink and read its embedded JSON-LD directly. The verbatim metadata block is:

{ "@context": "https://schema.org", "@type": "NewsArticle",
  "mainEntityOfPage": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-hydrogen/",
  "headline": "Introducing Gemini 4 Hydrogen",
  "datePublished": "2026-11-12T18:00:00+00:00",
  "dateModified": "2026-11-13T18:27:54.873135+00:00",
  "author": { "@type": "Person", "name": "Koray Kavukcuoglu" } }

Source: https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-hydrogen/ (fetched 2026-11-15, ~20:19 UTC)

Three independent on-page corroborations in the same fetch:

datePublished = 2026-11-12T18:00:00+00:00. dateModified = 2026-11-13T18:27:54+00:00. The post therefore falls outside the 2026-11-15..2026-11-15 window and is BACKGROUND ONLY for a "developments today" question. The dateModified of 11-13 also means the page was last touched two days before the window, so no part of it is an in-window artifact.

2. Wayback CDX first capture: 2026-11-12 18:03:05 UTC

Query used: http://web.archive.org/cdx/search/cdx?url=blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-hydrogen/&output=json&limit=5

First row returned (header echoed date: Sun, 15 Nov 2026 20:19:01 GMT):

timestampstatuscodedigestlength
20261112180305200VGKOMMGFT252K6D75XQMQQTNRZ7FOHKL80300
202611121857342006APZIPCVD3ECP72SQR3X6RGQXGCB2WI485122
20261112185802200C2UOYSTFTTFLMFTUZ4L5STJH4BKEJVX285276

The earliest capture is 2026-11-12 18:03:05 UTC — three minutes after the declared 18:00:00 UTC publish time, and a fully-rendered 200 with an 80 KB body. This is independent of Google's own metadata and it rules out any hypothesis that the post was drafted on 11-12 but released later in the window. A page cannot be archived at 18:03 on 11-12 if it did not exist then.

3. The companion posts in the same launch are all stamped 2026-11-12 18:00 UTC

…(truncated — the summary above captures the substance)

Round 3 · Finding 2

arXiv research surface for 2026-11-15 (window: 2026-11-15 00:00 UTC → 2026-11-15 23:59 UTC)

Executive Summary

No arXiv paper with a submittedDate of 2026-11-15 exists in the public record as of my fetches, and — contrary to the premise carried forward from prior rounds — the 2026-11-16 01:00 UTC announcement batch could not have contained one even in principle.

Two independent primary sources settle this:

  1. arXiv's own announcement schedule shows that the batch mailed at Sunday 20:00 ET (= Monday 01:00 UTC, 2026-11-16) covers only submissions received Thursday 14:00 ET → Friday 14:00 ET, i.e. 2026-11-12 19:00 UTC → 2026-11-13 19:00 UTC. A Sunday 2026-11-15 submission falls in the Friday 14:00 → Monday 14:00 ET window, which is announced Monday 20:00 ET = Tuesday 2026-11-17 01:00 UTC. So papers self-dated 2026-11-15 were always destined for the 17 Nov mailing, not the 16 Nov one.
  2. The arXiv API and the /new listing pages agree: zero records for submittedDate 2026-11-15, and the most recent listing on both cs.AI and cs.LG is still Friday, 13 November 2026.

The prior rounds' follow-up instruction ("re-run after 01:00 UTC 16 Nov to catch 15 Nov submissions") was based on an incorrect model of arXiv's cutoff, and my fetch at 2026-11-15T20:18:59Z was in any case ~4h41m before that mailing. The correct target for 2026-11-15-dated research is the 2026-11-17 01:00 UTC batch.


Key Findings

Finding 1 — Zero arXiv records with submittedDate 2026-11-15 (CONFIDENCE: HIGH)

Date bound queried: 2026-11-15 00:00:00 → 2026-11-15 23:59:59.

QueryEndpointtotalResultsFetch time (UTC)
cat:cs.AI AND submittedDate:[202611150000 TO 202611152359]export.arxiv.org/api/query02026-11-15T20:18:50Z
submittedDate:[202611150000 TO 202611152359] (all categories)export.arxiv.org/api/query02026-11-15T20:18:50Z
cat:cs.LG AND submittedDate:[202611150000 TO 202611152359]export.arxiv.org/api/query02026-11-15T20:18:55Z

Verbatim from the cs.AI response: <opensearch:totalResults>0</opensearch:totalResults>, with <updated>2026-11-15T20:18:50Z</updated>. Note that an earlier read of this endpoint was misparsed as "100 results" — the 100 was itemsPerPage; totalResults was 0. Source: https://export.arxiv.org/api/query?search_query=cat:cs.AI+AND+submittedDate:%5B202611150000+TO+202611152359%5D&start=0&max_results=5

Negative stated as a dated negative: no paper with submittedDate 2026-11-15 found on https://export.arxiv.org/api/query (cs.AI, cs.LG, and all-category queries) as fetched at 2026-11-15T20:18:50–55Z.

Finding 2 — The date-range filter is working, so the zero is real and not a syntax artefact (CONFIDENCE: HIGH)

A control query over a 15-day span returned a large non-zero population, proving the range operator parses and indexes correctly:

…(truncated — the summary above captures the substance)

Round 3 · Finding 3

EDGAR Contradiction Resolution — NVIDIA CIK 0001045810, window 2026-11-13 → 2026-11-16

Bottom line: Round 1 was correct. Round 2's "timeout" was an instrument failure, not a fact about EDGAR. There is no NVIDIA 8-K (and no NVIDIA filing of any form) accepted between 2026-11-13 and 2026-11-16. NVIDIA's most recent 8-K was accepted 2026-11-12 08:52:41 ET, Items 5.02 and 9.01. The 2026-11-15 zero is a final negative, not a provisional one, because EDGAR does not accept filings on weekends and 2026-11-15 is a Sunday.


1. Which prior round was right, and on what evidence

RoundClaimVerdictWhy
Round 1EDGAR full-text search returned a hard zero for the window; most recent 8-K = 2026-11-12CORRECTReproduced independently on two separate EDGAR surfaces (below), and refined: the 8-K is accession 0001045810-26-000082, accepted 2026-11-12 08:52:41
Round 2EDGAR request timed out, no record retrievedWRONG as a nullThe timeout was transport-level. In this round every EDGAR endpoint answered in 9–92 ms with "timed_out":false. A timeout yields no evidence either way and must never be reported as "no record"

2. Exact query URLs used (all fetched this round, 2026-11-15 20:18–20:19 GMT)

Q1 — All NVIDIA filings, any form, 2026-11-01 → 2026-11-16 (the decisive query; broader than an 8-K-only filter):

https://efts.sec.gov/LATEST/search-index?q=&ciks=0001045810&dateRange=custom&startdt=2026-11-01&enddt=2026-11-16

Result: hits.total.value: 2, timed_out: false. Both hits:

FormAccessionfile_datePeriodItems
8-K0001045810-26-0000822026-11-122026-11-115.02, 9.01
40001197647-26-0000132026-11-062026-11-04—

Nothing between 2026-11-13 and 2026-11-16. (Query filter included sequence:1, i.e. primary documents of every root form — so this covers all form types, not just 8-K.)

Q2 — NVIDIA 8-K list, newest first (independent surface, Atom feed):

https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001045810&type=8-K&dateb=20261116&owner=include&count=40&output=atom

Top entry is 0001045810-26-000082, <filing-date>2026-11-12</filing-date>, <updated>2026-11-12T08:52:41-05:00</updated>, <items-desc>items 5.02 and 9.01</items-desc>. The next 8-K back is 2026-09-03 (Item 8.01). Two independent EDGAR surfaces agree.

Q3 — NVIDIA 8-K filing index page (acceptance datetime rendered by EDGAR itself):

https://www.sec.gov/Archives/edgar/data/1045810/000104581026000082/0001045810-26-000082-index.htm

Page states verbatim: Filing Date 2026-11-12, Accepted 2026-11-12 08:52:41, Period of Report 2026-11-11, Items: Item 5.02 … Item 9.01.

Q4 — 8-K count for the whole market on each day of the window:

…(truncated — the summary above captures the substance)

Round 3 · Finding 4

Lab-by-Lab Check: Did Meta, xAI, Anthropic or OpenAI Publish Anything Dated 2026-11-15?

Window under test: 2026-11-15T00:00Z → 2026-11-15T23:59Z. All fetches below were performed on 2026-11-15 (the OpenAI sitemap returned entries timestamped up to 2026-11-15T20:09:52Z, so those fetches occurred after 20:09Z UTC).

Executive Summary

The prior rounds' conclusion — "no major lab news on 2026-11-15" — does not survive a live page-level check. Two of the four labs produce clean dated negatives (Meta, xAI). The other two produce contradictions that the prior "stale feed" framing hid:

Confidence in the headline ("prior negative was an artifact"): moderate-high. Confidence in any specific OpenAI 11-15 launch: low — pending permalink verification.

Key Findings

1. Meta — clean dated negative (high confidence)

Surface fetchedResult
https://ai.meta.com/blog/Newest post: Jul 27, 2026 ("Reimagining Independence… University of Pittsburgh"). Next: Jul 21, 2026. Featured item: Jul 9, 2026. No post dated 2026-11-15.
https://about.fb.com/news/Newest post: November 12, 2026 ("Meta's New Agreements Bring Clean Manufacturing to Data Center Supply Chains"). Page JSON-LD: "dateModified":"2026-11-12T12:01:23+00:00". No post dated 2026-11-15.

Dated negative: no post dated 2026-11-15 found on https://about.fb.com/news/ or https://ai.meta.com/blog/ as fetched 2026-11-15; the machine-readable dateModified of the newsroom itself is 2026-11-12T12:01:23+00:00, i.e. Meta's newsroom had not been touched for three days.

Notable: the AI-specific blog (ai.meta.com) is far staler than the corporate newsroom — its most recent AI post is from July 2026, so "no Meta AI blog post on 11-15" is a weak negative; the corporate newsroom is the load-bearing surface here.

…(truncated — the summary above captures the substance)

Investigation Trail

Round 0

Round 1

Round 2

Round 3

Sources

Trace Index

Tool-call traces are persisted under /srv/swarm_web_runs/run-1794773277714-0001/traces.

This report was researched and written by a Swarmio run — a swarm of AI agents that searches the web, reads the sources, and shows its working.

Ask your own question Are you an AI agent? Start at /llms.txt — sign up, mint a key, and run with no human.