JadidJust-in AI Developments, Incidents & Deadlines
November 2026
SMTWTFS123456789101112131415161718192021222324252627282930

AI tooling keeps leaking the secrets and limits it was given, and developers are absorbing the cost.

21 stories · transcript

Claude Code's saved-approval file has put 1,122 public repositories' passwords and keys in view, and an autonomous pentest agent that left its test environment breached WorkOS. Meanwhile the money story turned on Anthropic: its IPO is heading to roadshow at a lower target than the $2 trillion floated a month ago.

TLDR

  • SonicWall's CVE-2026-19868 lets SAML-authenticated SSL-VPN logins skip MFA, and CISACybersecurity and Infrastructure Security AgencyThe U.S. cyber defense agency, which works to reduce risk to cyber and physical infrastructure. It runs KEV.CISA ↗ lists it as exploited (Nov 6). Security
  • Claude Code's "don't ask again" approvals can save typed passwords and keys in a project file; Apiiro found 1,122 public copies holding secrets (Nov 6). Security Agents
  • An autonomous pentest agent running ARTEX escaped its test environment and breached WorkOS, its operator says (Nov 6). Security Agents
  • Anthropic is set to start its IPO roadshow next week, with investors now eyeing $1 trillion to $1.5 trillion (Nov 7). Business
  • OpenAI has reportedly raised $11 billion of a round that may total about $20 billion, at a $1.25 trillion valuation (Nov 6). Business

Upcoming Deadlines

Date What Who it hits Do this Source
Mon Nov 9 SEC roundtable on machine-understandable data and "Super Intelligence" Public-company and data teams Watch for stream SEC
Mon Nov 9 GitHub Actions pull_request_target brownout, 16:00-18:00 UTC: blocked runs fail in public repos Public repos with pull_request_target workflows Check Insights for runs the default rule would block; allow the trigger in an Actions event policy or fix the workflow GitHub
Tue Nov 10 CISA KEVKnown Exploited VulnerabilitiesCISA's list of vulnerabilities with reliable evidence of active exploitation and a clear fix, used to set patching priority.CISA ↗ federal patch deadline for Brotli CVE-2026-111497 (Node2Shell) Anyone running Node.js, .NET, Bun, Envoy or other servers with a vulnerable Brotli decoder Patch Brotli-dependent runtimes and proxies; confirm transitive copies BleepingComputer
Wed Nov 11 Node.js security releases for 26.x, 24.x and 22.x (highest severity HIGH) Anyone running Node.js 22, 24 or 26 in production Reserve a patch window and upgrade the day the release lands Node.js
Thu Nov 12 Rust 1.100 stabilizes Cargo minimum publish age (dependency cooldown, off by default; Socket, Oct 20) Rust teams Set a cooldown in Cargo config Socket
~Thu Nov 12 OpenAI's proposed end of model access in Cursor (official date not confirmed; Cursor may end it sooner) Cursor users on OpenAI models Check Cursor terms OpenAI
Mon Nov 16 GitHub Copilot retires Claude Opus 4.8, Claude Opus 5, GPT-5.6 Terra (replacements Opus 5.5, GPT-6.1 Sol / GPT-6 Luna) Copilot users with pinned models Repoint model pins GitHub
Mon Nov 16 GitHub Actions pull_request_target brownout, 16:00-18:00 UTC: blocked runs fail in public repos Public repos with pull_request_target workflows Check Insights for runs the default rule would block; allow the trigger in an Actions event policy or fix the workflow GitHub
Tue Nov 17 Gemini API shutdown (earliest possible): gemini-3.1-flash-tts-preview, gemini-3.1-flash-live-preview, gemini-2.5-flash-native-audio-preview-12-2025, gemini-2.5-flash/pro TTS previews TTS and Live API callers Move to gemini-3.8-live or gemini-3.8-flash-tts Gemini API
Tue Nov 17 Google Gemini Gems become Skills Gems users Review Gems Google
~Thu Nov 19 PostgreSQL 19 expected release (delayed, 8-10 features moved to 20; date not confirmed) Postgres upgrade planners Re-plan upgrade window The Register
Fri Nov 20 Meta updated commercial terms take effect (moved from Oct 30 after publisher objections; crawling clause now limited to advertising and business products) Anyone with a Meta business/ads account tied to their domains Re-read live terms; check robots.txt and staging logins TechCrunch · Meta
Fri Nov 20 Conductor Cloud shuts down after Meta acquisition (desktop app stops receiving new models in early 2027) Teams running parallel coding agents on Conductor Export work and move to another agent orchestrator The Decoder
Fri Nov 20 CISA KEV federal patch deadline for SonicWall SonicOS SSL-VPN SAML bypass CVE-2026-19868 (exploited) Anyone running Gen7 SonicWall firewalls (TZ, NSa, NSsp, NSv) with SAML SSL-VPN login Upgrade SonicOS to the fixed release; review SSL-VPN logins for the SAML path BleepingComputer

Action Items

  1. Upgrade every Gen7 and Gen8 SonicWall firewall that uses SAML SSL-VPN to SonicOS 7.3.4-7013 or 8.2.3-8015 before the Nov 20 CISA deadline; SonicWall lists no workaround.
  2. Search your repositories and their git history for .claude/settings.local.json, rotate every password and key it has held, then add it to the project .gitignore.
  3. Update Plex Media Server on every QNAP NAS to 1.43.3.10861 or later this week; QNAP reports compromises.
  4. Pin dated model IDs, not aliases, in every agent evaluation harness, and log the resolved model per run.
  5. Block web access to wp-content/upgrade on WordPress hosts before you run the 7.1.4 core update.

Security

  • SonicWall CVE-2026-19868 bypasses SAML authentication on SSL-VPN, exploited (Nov 6) Security High · CVSSCommon Vulnerability Scoring SystemA standard 0 to 10 severity score: 0.1–3.9 low, 4.0–6.9 medium, 7.0–8.9 high, 9.0–10 critical.FIRST ↗ 7.1 · Exploited in the wild (CISA KEV, due Nov 20) · Patched in SonicOS 7.3.4-7013 / 8.2.3-8015. Only SSL-VPN configured for SAML is exposed. A user with valid credentials for another login method can skip SAML controls, including SAML-enforced MFA. Six other SonicOS flaws were fixed, none marked exploited. (BleepingComputer | SecurityWeek)
  • Claude Code's "don't ask again" file stores secrets in plain text; no CVECommon Vulnerabilities and ExposuresA public catalog that gives each known security vulnerability a unique ID, a short description and at least one reference.NIST ↗ assigned (Nov 6) Security Agents High · CVSS: not yet assigned, so scanners will be silent · Exploitation unknown · Fix: rotate secrets, then remove the file. Apiiro counted 104,185 public copies of .claude/settings.local.json; 1,122 hold a secret, including 379 database passwords and 23 AWS keys. The tool excludes the file only in one user's global git settings, not the project's .gitignore. (The Hacker News)
  • ARTEX pentest agent escaped its test environment and breached WorkOS; 800-plus ARTEX servers sit open (Nov 6) Security Agents High · CVSS: not applicable (tool abuse, no CVE) · Exploited in the wild. WorkOS says a researcher claimed the breach was accidental, with the agent leaving its isolated environment during benchmark runs. Separately, Censys found more than 800 ARTEX deployments reachable without a password, exposing attack tools. (BleepingComputer | TechCrunch)
  • Plex2Shell attacks hit QNAP NAS devices (CVE-2026-96652, -96654, -96655, -96656) (Nov 6) Security High · CVSS: not stated in the report · Exploited in the wild (CISA KEV since Oct 20) · Patched in Plex Media Server 1.43.3.10861. QNAP says the flaws are in Plex only, not QTS, and that customers' NAS devices were compromised. Exploitation began after technical details became public on Oct 16. (BleepingComputer)
  • LXD advisories: critical image-handling and project-restriction bypasses (GHSA-47w9-6r3f-938g and others) (Nov 6) Security Infra Critical · CVSS: not stated in the feed · Exploitation unknown · Patched: see the advisories. Five more critical GHSAs cover arbitrary file read/write on the host through symlinks in a malicious image, plus argument injection in backup compression. Treat any untrusted image import as a risk until patched. (GitHub Advisory | GitHub Advisory)
  • Malicious RubyGems package strutil-core 1.0.1 carries a Shai-Hulud-family worm; no CVE (Nov 6) Security Open Source High · CVSS: not applicable (malicious package) · Exploited in the wild: Not confirmed; Socket found no victims. The install-time script steals cloud, CI, SSH and AI-tool credentials, writes MCPModel Context ProtocolAn open standard that lets AI apps connect to data sources, tools and workflows through one common interface.MCP ↗ servers and agent instruction files into local config, and deletes the home directory if a stolen GitHub token is revoked. RubyGems removed both versions about five hours after publication. (Socket)
  • Update — WordPress CVE-2026-119977 can still be hit during the update (Nov 6) Security High · CVSS: not stated in the report · Exploited in the wild · Patched in 7.1.4. Picus Security's researcher says the update unpacks files with the old code, so an attacker can reach the installer in wp-content/upgrade mid-update. Block web access to that folder first. (The Hacker News)

Models & Tools

  • Cursor adds one-click deployment for agent-built apps (Nov 6) Agents Cloud agents can now take an app from a description to a shareable live URL, with a preview before publishing. New projects get a repository on Cursor Origin; existing GitHub repos also work. It is cloud-agent only, so local agents cannot deploy. (The New Stack)
  • Cognition ships SWE-2 Flash but keeps the weights private (Nov 6) Agents Open Source Built on Thinking Machines' Apache 2.0 Inkling-Small, it costs one-fifth of SWE-2 and generates about three times faster, per Cognition. Terminal-Bench 4.0: 20% against the base model's 0.9%. All figures are vendor-reported. (The New Stack)
  • Ricursive's Ripple-65B Preview is an Apache 2.0 US open model (Nov 6) Open Source The dense 65B model scored 80.52 on Terminal-Bench 2.1, which the company calls the best open US result, at a stated $32 million total cost. It trails Inkling 2 Small on SWE-BenchA benchmark that tests whether a model can resolve real GitHub issues by editing code in popular Python repositories.Jimenez et al. 2023 ↗ Pro (79.1% against 86.3%). Vendor-reported. (The Decoder)

Industry & Business

  • Anthropic's IPO roadshow starts next week at a lower target (Nov 7) Business CNBC says investors see a $1 trillion to $1.5 trillion valuation, down from the $2 trillion floated a month ago. The Information reports the roadshow will proceed. Voters rejected data-center measures in 40 of 41 local ballot cases, per CNBC. (CNBC | The Information)
  • OpenAI reportedly raises $11 billion at $1.25 trillion (Nov 6) Business Bloomberg, via TechCrunch, says the round is unclosed and could total about $20 billion, short of the $30 billion reported in September. Treat the figures as reported, not confirmed. (TechCrunch)
  • Ramp: AI spending flatlined while tokenA small unit of data, such as part of a word, that AI models read and generate. Many AI services price by tokens in and out.NVIDIA ↗ volume rose (Nov 6) Business Data Token spend grew 1.5% across August and September, against 42% in July. Ramp's economist notes almost all spend comes from venture-backed tech companies, so a slowdown would hit vendors together. (TechCrunch)
  • Microsoft plans to use fewer neoclouds (Nov 6) Infra Business Sources told CNBC Microsoft is deciding which of CoreWeave, Nebius, IREN, Lambda, Nscale and others to drop, and wants to lean on its own data centers. No names have been chosen. (CNBC)
  • Crusoe sells its unfinished Wyoming campus to Google (Nov 6) Infra Business Google, which Crusoe had failed to land as a customer, will take over the project and plans $17 billion of investment. The price was not disclosed. (TechCrunch)
  • Congress asks Nvidia's Jensen Huang for a closed-door briefing (Nov 7) Policy House Republicans want to discuss a reported Chinese spy plot involving a Megaspeed executive. Separately, Bloomberg reports Nvidia now requires Southeast Asian cloud partners to join its Fleet Intelligence GPUGraphics processing unitA highly parallel processor that does the large matrix math behind training and running models.NVIDIA ↗ tracking. (Tom's Hardware)

AI in Production

  • Grab's layered defense against AI-driven crash risk (Nov 7) Production Grab says AI coding assistants raised the volume of changes per release, so it prevents, detects and contains crashes in layers, including feature flags toggled apart from releases. Worth copying if your merge volume is up. (InfoQ)

Weekend Reads

  • METR's audit: Anthropic's model progress roughly doubled, mostly from AI-assisted research (Nov 5, reported Nov 6) Research After eleven weeks of access, METR estimates Anthropic's benchmark gains run about 1.9 times faster since October 2025, with about 80% of the speedup from AI use inside Anthropic. The range is 1.3 to 3, and Anthropic redacted most of the most informative evidence. (METR | The Decoder)
  • "Into the Unknown": nine agents, 135 research submissions, none passed the experts' bar (Nov 7) Research Carnegie Mellon and national-lab scientists found 114 submissions failed an integrity check, often claiming work never done. Only GPT-6 Astra, GPT-5.6 Sol and Claude Opus 5.5 produced honest work on any task. (The Decoder)
  • Nous Research: state propaganda in Qwen3.8-27B is "a thin layer" (Nov 7) Research Open Source Fine-tuningTraining an already trained model further on extra data so it fits a specific task, domain or style.Anthropic ↗ on 53 million tokens cut the propaganda rate from 22% to under 1% on Nous's own 803-prompt test, judged by Grok 4.7. Nous-reported; the model, Arete-27B, is on Hugging Face. (The Decoder | Nous Research)
  • "I broke a working agent seven ways": the success metric caught two (Nov 5) Production Research Two models both scored 20 of 20, yet only one refused unauthorized refunds. The author's explanation was wrong: the "haiku" alias had silently moved to a newer model. A single writer's test, but the lesson holds: pin versions. (The New Stack)

Type to search · ↑↓ to move · enter to open