Case file — Lead incident: The world's most valuable chipmaker is on a ransomware gang's leak site — and nobody can yet say whether the gang actually has anything. The Gentlemen, a ransomware-as-a-service operation that emerged in mid-2025, added Nvidia to its dark-web extortion portal on November 18, 2026, with a countdown of roughly 234 hours — just under ten days — before it says it will publish stolen data. The listing shows Nvidia's name, logo, stock ticker and a $303 billion revenue figure, and the words "DATA INFO – SOON." It shows no sample files, no file list, no data volume, no intrusion date, and no description of what was supposedly taken. Nvidia has not confirmed a breach. The company told PCMag it is "investigating an incident that may have affected a third-party partner" and has "found no indication that Nvidia systems or networks were affected." Trackers classify the entry as pending — the claim exists, and nobody has validated it. Status: claim unproven; countdown runs to about November 28.
Incident timeline
| Date | Development |
|---|---|
| October 6, 2026 | The Gentlemen lists Jentech Precision Industrial, a supplier of heat spreaders and cooling parts for Nvidia's AI chips, on its leak site. |
| October 27, 2026 | The Gentlemen lists FPT Software, a Nvidia technology partner in Southeast Asia, on its leak site. |
| November 18, 2026 | The Gentlemen adds Nvidia — and AARP — to its leak site; Nvidia's entry carries a countdown of about 234 hours and the label "DATA INFO – SOON." |
| November 18, 2026 | F5 ships nginx 1.30.7 and 1.31.9, fixing five vulnerabilities including the decade-old HTTP/2 flaw CVE-2026-123007; Rapid7 publishes a proof of concept. |
| November 18, 2026 | CISA adds four vulnerabilities to its Known Exploited Vulnerabilities catalog — Synology DSM, Apache Axis, REDCap and LiteLLM — with three carrying a November 21 deadline. |
| November 18, 2026 | Rhysida claims it exfiltrated 1.18 TiB — 2,773,480 files — from Dutch heavy-machinery dealer Van der Spek Vianen. |
| November 18, 2026 | Settra lists Biokinetics and Associates of Ottawa (100 GB) and Powers & Sons of the United States (50 GB); Pixelleaks claims Permit Pros and sets a November 22 deadline. |
1. Nvidia named on ransomware leak site — the claim is unproven and the evidence is empty
The biggest cybersecurity headline of the day is a claim, not a confirmed breach, and the distinction matters. On November 18, 2026, the ransomware group calling itself The Gentlemen added Nvidia to its dark-web leak site. The entry carries a countdown of roughly 234 hours — just under ten days — expiring around the morning of Saturday, November 28, UTC, and promises a data release under the label "DATA INFO – SOON." That is the extent of it. There are no sample files, no file tree, no screenshots, no stated data volume, no intrusion date, and no description of which part of Nvidia was supposedly hit.
Nvidia has not confirmed a breach. The company told PCMag it is "investigating an incident that may have affected a third-party partner" and has "found no indication that Nvidia systems or networks were affected." That wording leaves open the possibility that the claim traces back to a supplier rather than to Nvidia itself — and two of Nvidia's suppliers are already on the same leak site. Jentech Precision Industrial, which makes heat spreaders and cooling parts for Nvidia's AI chips, was listed on October 6; FPT Software, a Nvidia technology partner in Southeast Asia, was listed on October 27. Neither listing proves the Nvidia entry is supplier data either. What is established is only that the gang posted the name and started a clock.
The Gentlemen emerged in mid-2025 and operates as a ransomware-as-a-service crew: affiliates carry out intrusions and keep most of the ransom, with the operators taking a cut for the infrastructure. Ransomware trackers put the group at 1,123 claimed victims across 95 countries, and Check Point's October 2026 threat report named it the most prevalent ransomware group that month, accounting for 13% of published attacks. But only a small fraction of its claims — roughly 93 to 95 — have been confirmed through victim statements or filings, and researchers say the group has exaggerated and misattributed claims before. The group also suffered an internal data leak of its own in May 2026. A countdown is a deadline the gang set, not evidence of a breach, and until the group publishes files — or a victim confirms an incident — the honest summary is one line: a ransomware gang says it has something, and has shown nothing.
Investigation notes
What is confirmed: the leak-site entry and its November 18 timestamp (logged by the ransomware.live tracker at 09:22 UTC and by Ransomnews at 08:48 UTC); the countdown of roughly 234 hours; the absence of samples, file lists or data volume; and Nvidia's statement to PCMag that it is investigating a possible third-party partner incident with no indication its own systems were affected. The two supplier listings — Jentech on October 6 and FPT Software on October 27 — are confirmed entries on the same leak site.
What is not known: whether Nvidia or any supplier was actually breached, what data — if any — the gang holds, when any intrusion occurred, and what happens when the countdown ends. The gang's own history of exaggerated claims means the listing cannot be treated as proof.
What to watch: the countdown expires around November 28. If the group publishes files, the contents — not the listing — will establish what was actually taken and from whom. Nvidia suppliers and partners should check whether their own data appears in any future release.
2. nginx2root: a ten-year-old HTTP/2 flaw lets unauthenticated attackers run code on servers
F5 disclosed five vulnerabilities in nginx on November 18, and the worst of them has been sitting in the code since HTTP/2 support arrived in 2015. CVE-2026-123007 is an integer underflow in nginx's HTTP/2 flow-control handling: when a stream is cancelled with data still queued, the server adds unused flow-control allowance back to its send window without capping it, and a later unsigned subtraction underflows — corrupting memory in the worker process. An unauthenticated remote attacker with a reachable HTTP/2 port can leak memory addresses, crash the worker, and — where ASLR is disabled or can be bypassed — run code with the worker process's privileges. Wiz Research confirmed remote code execution, including an ASLR bypass, against official nginx 1.31.7 and 1.31.8 builds serving ordinary static files. Rapid7 has published a proof of concept.
The five flaws are CVE-2026-123007 (HTTP/2 flow control), CVE-2026-123633 (HTTP/2 proxying), CVE-2026-123634 (proxied server names built from variables), CVE-2026-121810 (HTTP/3) and CVE-2026-113425 (mail proxy). F5 rates each 8.1 under CVSS 3.1 and 9.2 under CVSS 4.0. The fixes shipped November 18 in nginx 1.30.7 (stable) and 1.31.9 (mainline). The researchers who found them read like a roll call of independent discovery: Xint (the autonomous platform from Theori), Wiz, Rapid7, Ikotas Labs and Metnir all reported the flaws, with Xint's automated harness finding four of the five in about 90 minutes of scanning roughly 50,000 lines of code. Xint calls the set "nginx2root"; Ikotas Labs calls the HTTP/2 flaw "NGINX2Shell."
The exposure is enormous but needs careful reading. Xint reports that, as of November 10, Shodan returned 56.7 million hosts running nginx, of which 11.94 million reported an identifiable version — and 98.9% of those fell within the range affected by CVE-2026-123007. Those figures describe detected hosts and reported version strings, not confirmed vulnerable servers: they do not account for backported distribution fixes, and they do not confirm that HTTP/2 is enabled or reachable on each host. Turning off HTTP/2 covers only two of the five flaws. No exploitation in the wild has been reported, and neither flaw is in CISA's Known Exploited Vulnerabilities catalog — but with a public proof of concept and a fix that is one upgrade away, the patching window is now.
Investigation notes
Official anchor: F5's nginx security advisories of November 18, 2026, and the CVE-2026-123007 record, which credits Rapid7, Wiz Research, Xint and Ikotas Labs as independent reporters.
What is confirmed: the five vulnerabilities, the affected versions (nginx 1.9.5 through 1.31.8 for the HTTP/2 flaw), the fixed releases 1.30.7 and 1.31.9, Wiz's confirmed remote code execution with an ASLR bypass, and Rapid7's published proof of concept. F5 disclosed ten nginx vulnerabilities in total on November 18.
What is not known: whether any attacker has exploited the flaws — none has been reported, and the flaws are not in CISA's KEV catalog. Xint also reported chaining CVE-2026-123007 with a separate Linux kernel SCTP flaw to reach root on the host; that chain has not been independently confirmed.
What defenders should do: upgrade to nginx 1.30.7 or 1.31.9 now, or apply the distribution's backported fix; where that is not possible today, disable HTTP/2 and HTTP/3 on internet-facing servers — which covers two of the five flaws — and rebuild container images carrying vulnerable nginx builds.
3. CISA adds four exploited flaws to KEV — three with a three-day deadline
CISA added four vulnerabilities to its Known Exploited Vulnerabilities catalog on November 18, citing evidence of active exploitation, and gave federal civilian agencies until Saturday, November 21, to deal with three of them. The additions are Synology DiskStation Manager (CVE-2024-10441), Apache Axis 1.x (CVE-2026-90436), Vanderbilt University's REDCap research data platform (CVE-2026-90817), and BerriAI's LiteLLM AI gateway (CVE-2026-40217). The first three are all rated 9.8 and all three are flagged for forensic triage — a check of whether the system was already compromised — under Binding Operational Directive 26-04. LiteLLM, rated 5.9, is due December 2 and is not flagged for triage.
The details make the short deadlines understandable. CVE-2024-10441 is a 2024 flaw in DSM's system plugin daemon that lets a remote attacker run code without logging in; the fix has been available for a year in DSM 7.2.2-72806-1, 7.2.1-69057-6, 7.2-64570-4 and later, and BeeStation OS 1.1-65374. CVE-2026-90436 is an unsafe-reflection flaw in Apache Axis 1.x: a crafted SOAP fault can make Axis load and create a class of the attacker's choosing before any authentication runs — and Apache does not expect to release a fixed Axis 1.x, recommending migration to another SOAP engine or applying a published patch to a custom build. CVE-2026-90817 is a code-injection flaw in REDCap. CVE-2026-40217 is a sandbox escape in LiteLLM's custom-code guardrail test endpoint that lets someone with proxy-admin rights run code; the default Docker image runs as root. CISA marks known ransomware use as "unknown" for all four and does not say who is exploiting them.
The additions follow CISA's November 17 KEV entry for the Perforce Puppet Enterprise deserialization flaw CVE-2026-115731, which this journal's hunting series covered on November 18 — that flaw lets an attacker obtain the Puppet server's certificate-signing key and push trusted certificates and altered configuration to managed machines, with a federal due date of December 1. With Java4Shell still unpatched in many estates and the nginx flaws disclosed the same day, the catalog now carries a cluster of actively exploited and weaponizable flaws with deadlines inside the week.
Investigation notes
Official anchor: CISA's Known Exploited Vulnerabilities catalog, version 2026.11.18, and the vendor advisories from Synology, Apache and BerriAI.
What is confirmed: the four additions, their due dates (November 21 for Synology, Axis and REDCap; December 2 for LiteLLM), the forensic-triage flags on the first three, and the CVSS scores — 9.8 for Synology, Axis and REDCap, 5.9 for LiteLLM. Apache's position that no Axis 1.x fix is expected is confirmed in its advisory.
What is not known: who is exploiting the four flaws, how many organizations have been hit, and whether any of the exploitation is tied to ransomware. CISA lists known ransomware use as unknown for all four.
What defenders should do: patch Synology DSM, REDCap and LiteLLM to the fixed releases before the deadlines; for Apache Axis 1.x, plan migration off the library or apply Apache's patch to a custom build, and treat any exposed Axis instance as potentially already compromised until triage says otherwise.
4. Rhysida claims 1.18 TiB stolen from Dutch heavy-machinery dealer Van der Spek Vianen
The Rhysida ransomware group says it exfiltrated 1.18 TiB of data — 2,773,480 files — from Van der Spek Vianen B.V., a Netherlands-based dealer of Komatsu machinery and tower cranes. The claim, logged on November 18, describes bookkeeping records spanning 2015 to 2026, including general ledger rows, details for 5,583 customers and 5,044 vendors, and 4,672 vendor IBANs. The haul reportedly also includes live website database passwords for domains including vanderspek.nl and used-komatsu.nl, and a NAV license key. The scale of the file count and the specificity of the records — eleven years of bookkeeping, thousands of customer and vendor entries — point to a deep intrusion rather than a smash-and-grab, though the claim itself remains the group's word until samples or a victim statement confirm it.
Rhysida is one of the more established ransomware operations still active, and its listings typically carry detailed inventories of the data it says it holds. For a mid-sized dealer, the exposure is severe on its own terms: customer and vendor records, bank account identifiers and live database credentials are the raw material for fraud, impersonation and follow-on attacks against the company's partners. Organizations that trade with Van der Spek Vianen should treat the claim as credible enough to watch for: monitor for phishing using the company's name, verify any payment-detail changes through a known contact, and rotate credentials shared with the dealer.
Investigation notes
What is confirmed: the leak-site listing dated November 18, 2026, as indexed by the ransomware.live tracker, with the stated volume of 1.18 TiB and 2,773,480 files, and the categories of data described in the listing.
What is not known: whether the data is authentic or complete, whether systems were encrypted as well as exfiltrated, and whether Van der Spek Vianen has confirmed the incident — no victim statement has been reported.
What defenders should do: customers and vendors of the company should treat payment-detail and credential-change requests with extra scrutiny, and any organization in the same sector should check for the exposure of its own data in the listing if samples appear.
5. Ransomware roundup: Settra hits two, Pixelleaks claims a permit platform
Three more ransomware claims landed on November 18, all from groups working the same playbook of listing first and negotiating under a deadline. Settra, a newer group, published two victims. Biokinetics and Associates Ltd., a Canadian company based in Ottawa, was listed with 100 GB of exfiltrated data, including internal financial statements and third-party medical records the company held as part of litigation discovery; the attack is estimated to have taken place on October 21, 2026. Powers & Sons, a United States manufacturing company, was listed the same day with 50 GB taken in an attack estimated at October 22, 2026, including the home addresses and personal phone numbers of the company president, a co-owner and a vice president.
Separately, the Pixelleaks group claimed Permit Pros LLC, a United States license-management platform serving national restaurant and service chains, saying it obtained complete backups of all 21 client databases hosted by the platform — 103,196 verified data rows containing license and permit records, including records for tenants such as Chuck E. Cheese. The group set a deadline of November 22, 2026, at 13:11 UTC. The report itself cautions that the claim is emerging and should be treated with care until independently verified. None of the three victims has publicly confirmed an incident, and no data samples have been published for any of them.
Investigation notes
What is confirmed: the three leak-site listings dated November 18, 2026, as indexed by ransomware trackers, with the stated volumes (100 GB, 50 GB, and 103,196 rows across 21 databases) and the Pixelleaks deadline of November 22 at 13:11 UTC.
What is not known: whether the data is authentic, whether any of the three organizations has confirmed an incident, and what was taken beyond the categories the groups describe. The Permit Pros claim is explicitly flagged by its own source as unverified.
What to watch: the Pixelleaks deadline on November 22 — if data is published, its contents will show whether the claim reaches the client databases the group describes.
Sources
- NVIDIA ransomware claim: listing posted, no proof shown — Candid Dragon
- 2shell2much: NGINX2Root vulnerabilities potentially expose more than 10m websites — Cyber Daily
- Remote code execution in NGINX HTTP/2 (CVE-2026-123633, CVE-2026-123007) — nginx2shell.com
- CISA gives agencies three days to fix exploited REDCap, Synology and Apache Axis flaws — MLQ
- Rhysida Ransomware Group Exfiltrates 1.18 TiB of Data from Van der Spek Vianen B.V. — QPulse
- Settra Ransomware Group Exfiltrates 100GB of Data from Biokinetics and Associates Ltd. — QPulse
- Pixelleaks Ransomware Group Claims Data Exfiltration from Permit Pros LLC — QPulse
