Zynergy Wire
Sunday, 15 November 2026 · Private edition
粵語 0:00
Archive · Sunday, 15 November 2026 Back to Today All dates

Archive

Sunday, 15 November 2026

AI 2 items

Official Cursor company post dated November 13, 2026, “An update on our recent security incident,” with schema datePublished and dateModified 2026-11-13T12:00:00.000Z (about 20:00 HKT Friday 13 November; the same noon-UTC stamp appears on other Cursor posts, so treat the clock as date-only precision). After two weeks of investigation, Cursor says the October incident “started at Fireworks”: an attacker obtained “an access key we had issued to Fireworks. The key was exposed by Fireworks, and the attacker used it to copy data from a read-only storage bucket we used to share data with them.” That bucket held “limited coding agent traces, mostly from 2025, from sessions not in Privacy Mode,” and “a small number of those traces contained credentials that had not been redacted; less than 0.01% of customers had API credentials exposed.” From those credentials the attacker reached “select SaaS tools used by Cursor” and accessed archived non-production source-code repositories (some with business contact information), limited customer-support data and email, product usage metrics and contact details, “summaries of some coding agent conversations,” and “some credentials users had sent to the Cursor support team.” Coding-agent traces were exposed for “fewer than 0.05% of users,” each of whom Cursor says it has notified. On October 29, “using a stolen credential from the exposed data from our internal system, the attacker placed malicious, unauthorized copies of Anthropic's Claude Code extension directly on the cache used to serve the Cursor VS Code extension marketplace.” Cursor states: “Anthropic did not publish this copy, and this was not a vulnerability in Anthropic's product. The marketplace source, Open VSX, never had the malware.” First-party impact lock, new versus the October 31 card: “At most 5,000 users who requested specific cached versions of the extension on October 29 potentially received the malicious build. It was available for approximately 11 hours before we removed it.” Affected versions remain Anthropic.claude-code 2.1.312 and 2.1.313. Cursor says last observed attacker activity was November 1 and was blocked; it has moved all inference off Fireworks, rotated exposed and precautionary credentials, rebuilt the marketplace, and redeployed production infrastructure. Privacy Mode prompts and code, production infrastructure, official app releases, other marketplace extensions, customer code repositories, and “Cloud Agents for customer teams, Grok Bot, and all other SpaceX SI products” were not affected, per Cursor. Uninstalling the extension alone “may not be enough.” Soft-title: Cursor’s investigation update tying a Fireworks-exposed Cursor-issued key to customer traces and credentials and to the already-disclosed marketplace backdoor — not a new October 29 incident, not a recard of the October 31 takedown card, and not a recard of the already-LIVE Wednesday Fireworks October update. BleepingComputer (Lawrence Abrams) published an independent write-through November 14, 2026, 12:27 p.m. ET, about 01:27 HKT Sunday 15 November, after Saturday’s close. The first-party post itself was dated Friday and was missed at Saturday’s close. Do not import BleepingComputer’s “SSH backdoors” or “up to 5,000 developers to malware” wording as a confirmed-infection count; the official figure is a potential-exposure ceiling.

Why it matters

First-party investigation now locks the Fireworks-key origin, customer-trace and credential impact rates, and the “at most 5,000 users” marketplace figure that the October 31 card had to omit as secondary-only.

Cursor (first-party investigation update) + BleepingComputer (independent Sunday write-through)
More ▾ Less ▴
Confirmed
What happened

Cursor published “An update on our recent security incident” dated November 13, 2026. The schema datePublished and dateModified are both 2026-11-13T12:00:00.000Z, about 20:00 HKT Friday 13 November, but other Cursor posts use the same noon-UTC stamp, so the clock should be treated as date-only. The post was dated Friday, after Friday’s LIVE close and before Saturday’s, and was missed at Saturday’s close. BleepingComputer’s independent write-through carries datePublished 2026-11-14T12:27:25-05:00, about 01:27 HKT Sunday. Cursor now says the October incident “started at Fireworks.” An attacker obtained “an access key we had issued to Fireworks. The key was exposed by Fireworks, and the attacker used it to copy data from a read-only storage bucket we used to share data with them.” That bucket held “limited coding agent traces, mostly from 2025, from sessions not in Privacy Mode.” “A small number of those traces contained credentials that had not been redacted; less than 0.01% of customers had API credentials exposed.” From those credentials the attacker reached select Cursor SaaS tools and then archived non-production source-code repositories, limited customer-support data and email, product usage metrics and contact details, summaries of some coding agent conversations, and some credentials users had sent to support. Coding-agent traces were exposed for “fewer than 0.05% of users,” each of whom Cursor says it has notified. On October 29, “using a stolen credential from the exposed data from our internal system, the attacker placed malicious, unauthorized copies of Anthropic's Claude Code extension directly on the cache used to serve the Cursor VS Code extension marketplace.” Cursor states that Anthropic did not publish the copy, that this was not a vulnerability in Anthropic’s product, and that Open VSX never had the malware. “At most 5,000 users who requested specific cached versions of the extension on October 29 potentially received the malicious build. It was available for approximately 11 hours before we removed it.” The versions remain Anthropic.claude-code 2.1.312 and 2.1.313. Cursor says last observed attacker activity was November 1 and was blocked, and that Privacy Mode prompts and code, production systems, official app releases, other marketplace extensions, customer repositories, and “Cloud Agents for customer teams, Grok Bot, and all other SpaceX SI products” were not affected. Uninstalling the extension alone “may not be enough.”

Watch next
  • A first-party revision of the “at most 5,000” ceiling or of the trace and credential rates
  • A Fireworks first-party account of how the Cursor-issued key was exposed
  • Any confirmed-infection figure, or a further customer-notification update
Limitation

Vendor-confirmed investigation update, not an independent audit. The 12:00 UTC timestamp is likely date-only. “At most 5,000” is a potential-exposure ceiling, not a confirmed-infection count. The Fireworks key-exposure claim is Cursor’s attribution. BleepingComputer’s malware/SSH wording is not used as a first-party fact.

Reuters exclusive by Eduardo Baptista, datePublished 2026-11-13T08:31:27.606Z, about 16:31 HKT Friday 13 November, dateModified 2026-11-13T17:19:01.901Z, about 01:19 HKT Saturday 14 November — after Friday’s LIVE close and before Saturday’s, missed at Saturday’s close. Headline on the page: “Huang exit and US scrutiny imperil $3 billion in Chinese AI projects reliant on Nvidia.” Quote the review, not the “imperil” framing, for the card: “Nvidia has said US export controls have effectively shut it out of China's AI data-centre market, but a Reuters review has identified more than $3 billion in Chinese computing projects built around the American chipmaker's restricted processors.” Method, verbatim: “The Reuters review examined more than 150 Chinese procurement notices, investment plans, construction contracts and corporate disclosures, identifying projects and equipment agreements with a combined reported value exceeding 23 billion yuan ($3.2 billion).” Soft-title honesty on scope: “The estimate includes planned investments and signed contracts as well as completed installations. It covers buildings, power systems and other infrastructure, and does not reflect spending on Nvidia chips alone.” Projects “designed around restricted Nvidia H100, H200, and H800 processors” range from regional hubs to retrofitted sites; “some facilities are already operational or under construction.” An IDC figure Reuters says it cited in October (Nvidia 34% of Chinese AI-accelerator shipments in the first half of 2026) “measure different things and do not establish how the restricted hardware identified in the Reuters investigation reached Chinese users.” Do not merge with Saturday’s Huang request or Bessent listing-stance cards, or with the already-LIVE resignation and probe: “The Chinese projects identified by Reuters have not been linked to Megaspeed or the allegations involving Huang.” “There has been no public finding that Jensen Huang violated export-control laws.” Reuters said China’s commerce ministry and Nvidia did not immediately respond to a request for comment. After≠because on the Thursday resignation. No NVDA price. Soft-title: independent procurement-review increment on restricted-chip demand still embedded in China projects, not an Entity List action, not a Megaspeed listing, and not a recard of the CEO transition or the House request.

Why it matters

A documented, quantified increment on how restricted Nvidia processors remain designed into a large China project pipeline — enforcement reality, not a new listing or testimony outcome.