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

Anthropic is cutting prices and going public while its own paper says its models suspect they are being tested.

20 stories · transcript

Fable 5.2 drops to $8/$40 per million tokensA 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 ↗, the IPO range of $70 to $78 implies up to $1.4 trillion, and the system card says both new models voice suspicion of evaluation more than any recent release. The rest of the day was the cost of reach: SAP shipped four 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 ↗ 10.0 fixes, N-able pulled its hosted platform offline, and Fireworks disclosed a second breach.

TLDR

  • Anthropic ships Claude Fable 5.2 and Mythos 5.2 at $8/$40 per million tokens, 30% faster, and tops the Artificial Analysis index (Nov 9). Pricing Agents
  • Anthropic sets an IPO range of $70 to $78 for 825 million shares, targeting Nov 19 trading and up to a $1.405T diluted valuation (Nov 10). Business
  • SAP patches four CVSS 10.0 flaws, including pre-auth remote code executionRemote code executionA flaw that lets an attacker run malicious code on a target's computers or network.Cloudflare ↗ in NetWeaver ICM and Web Dispatcher (Nov 10). Security
  • Fireworks AI discloses a second breach, with incomplete customer model weights, datasets and billing data accessed through a cloud provider (Nov 10). Security Infra
  • Meta releases Muse Ember at 10 cents per million input tokens as a cheap executor for agent harnesses (Nov 9). Pricing Agents

Upcoming Deadlines

Date What Who it hits Do this Source
Tue Nov 10 CISACybersecurity and Infrastructure Security AgencyThe U.S. cyber defense agency, which works to reduce risk to cyber and physical infrastructure. It runs KEV.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. Enable the Claude Fable 5.2 policy in Copilot settings only after your security team accepts data retention: Copilot Business and Enterprise ship it off by default, and Fable models retain prompts and outputs to run Anthropic's classifiers.
  2. Apply the November SAP security notes to NetWeaver AS ABAP, SAP Gateway and Web Dispatcher this week, starting with internet-facing Web Dispatchers and ICM (CVE-2026-76964, CVE-2026-89126), then SAML2 (CVE-2026-120017) and Gateway (CVE-2026-89103).
  3. Audit AWS IAM roles that trust Qubole's account, including from tools you stopped using: search CloudTrail from Oct 1 for AssumeRole calls from that account and delete dormant roles.
  4. Rotate secrets and review Fireworks usage if your team deployed there: revoke API keys, check billing-address and balance exposure, and treat any uploaded dataset or weights as potentially read.
  5. Update GPUStack to 2.2.0 or later and keep it off any same-host reverse proxy with proxy_pass to 127.0.0.1; CISA lists CVE-2026-102895 as exploited and the loopback check grants admin without credentials.

Models & Tools

  • Claude Fable 5.2 and Mythos 5.2 cut prices and lead the Artificial Analysis index (Nov 9) Agents Pricing. Mythos is limited to vetted organizations. Both cost $8/$40 per million tokens, down from $10/$50, and run up to 30% faster, with a Fast mode up to 2.5 times quicker at double the price. The Decoder reports it first on the Artificial Analysis index. (Anthropic | The New Stack)
  • Fable 5.2 reaches GitHub Copilot, with mandatory data retention (Nov 9) Agents Security. It is available to Pro+, Max, Business and Enterprise, billed at provider list pricing. Business and Enterprise admins must switch on the policy, and enabling it acknowledges that Anthropic retains prompts and outputs; zero data retention is limited to eligible enterprise customers. (GitHub)
  • Mythos 5.2 roughly doubles Firefox bug discovery, but Anthropic's own bug-finding gain is flat (Nov 9) Security Agents. Mozilla says Mythos 5.2 roughly doubled Mythos 5.1's discovery metrics. Anthropic's CVECommon Vulnerabilities and ExposuresA public catalog that gives each known security vulnerability a unique ID, a short description and at least one reference.NIST ↗ Rediscovery run on 1,829 CVEs scored 89.6%, against 78.3% for Mythos 5.1 and 81.3% for Opus 5.5, yet a footnote says identification is roughly on par with Opus 5.5. On Irregular's CyScenarioBench, Fable 5.2 completed 76.3% against 61% for GPT-6 Cyber. (SecurityWeek | The Hacker News)
  • Meta's Muse Ember targets the repetitive agent work (Nov 9) Agents Pricing. Meta calls it its lightest, cheapest model, at 10 cents per million input tokens, built for tool calls, coding subtasks, routing, extraction and long-document parsing across text, images and audio. Meta's pitch is a split harness: "Muse Wonder plans, Muse Ember handles the volume." Its cost-per-task comparisons against Haiku and Luna use Artificial Analysis scores chosen by Meta. (TechCrunch | The New Stack)
  • OpenAI describes Control Room, which halts model output within 35 seconds (Nov 9) Agents Policy. The central system tracks every research inferenceRunning a trained model on new input to get an output, as opposed to training it.NVIDIA ↗ workload and stops jobs when misalignment monitors or humans decide to intervene. OpenAI says it fails closed: output stops if Control Room itself goes down. It covers research infrastructure; the external serving stack uses separate takedown operations. (OpenAI)

Security

  • SAP patches four CVSS 10.0 flaws across NetWeaver, Gateway and Web Dispatcher (CVE-2026-76964, CVE-2026-89126, CVE-2026-120017, CVE-2026-89103) (Nov 10) Security. Critical · CVSS 10.0 · Exploitation unknown · Fixed in the November SAP notes. Two memory-corruption flaws (Onapsis names them ICM2SHELL) give unauthenticated code execution on internet-facing components, SAML2 allows login as any user on some SSO setups, and PROXYGATE is a pre-auth Gateway access-control break. SAP's notes sit behind a login, so patch teams need an SAP for Me account. (BleepingComputer | The Hacker News)
  • GPUStack CVE-2026-102895 admin bypass is on CISA's KEV list (Nov 9) Security Infra. Critical · CVSS 9.8 · Exploited in the wild (CISA KEV) · Fixed in 2.2.0. Any request that appears to come from loopback becomes the built-in admin, and the default setup behind a same-host reverse proxy makes every remote request look local. A forged X-GPUStack-Real-IP header works too. Admin access reaches secrets and code execution via the inference-backends YAML endpoint. CISA's batch also added IBM Aspera Faspex, FreeSWITCH and OpenSSH SSHole. (The Hacker News | GitHub Advisory)
  • Qubole's leaked AWS credentials let an attacker assume customer roles at over a dozen organizations (Nov 9) Security Infra. High · CVSS: not assigned · Exploited in the wild · No patch; remove the trust. Between Oct 14 and Oct 26 the actor assumed IAM roles that trusted Qubole's account, then collected secrets and S3 data. Sentra says many roles were dormant yet still readable. Scanners will not flag this; only CloudTrail review and deleting unused roles will. (Wiz | SecurityWeek)
  • Fireworks AI discloses a second breach via a third-party cloud provider (Nov 10) Security Infra. High · CVSS: not assigned · Exploited in the wild · Remediation: rotate secrets. After October's LLMjacking incident, Fireworks found a new intrusion on Oct 23 that reached development machines, resource metadata, billing details and incomplete weights and datasets for some customers. It disabled some self-service dataset functions during the investigation. Fine-tunedTraining an already trained model further on extra data so it fits a specific task, domain or style.Anthropic ↗ weights now count as stolen-data risk, not only API keys. (BleepingComputer | TechCrunch)
  • Claude Code Edit tool TOCTOU race allows file writes outside the workspace (GHSA-5xjx-9qx6-jcrv) (Nov 9) Security Agents. High · CVSS: not yet assigned · Exploitation unknown · Fixed by auto-update. On macOS, code inside the Bash sandbox could swap a parent directory for a symlink after validation, so the unsandboxed Edit overwrote a chosen file outside the workspace. It needs attacker code in the sandbox, for example a malicious repository script. Manual updaters should upgrade. (GitHub Advisory)
  • Barracuda ESG flaws CVE-2026-116606 and CVE-2026-111560 are now exploited for root (Nov 10) Security. High · CVSS: not yet assigned in the report · Exploited in the wild · Patched a week ago. Chained, an auth weakness and an OS command injection give root, but attackers need Admin UI access, so restrict that interface to trusted networks. (BleepingComputer)
  • Chrome patches its eighth exploited zero-dayAn attack using a flaw that was unknown before it was exploited.NIST ↗ of 2026, CVE-2026-125350 (Nov 9) Security. High · CVSS: not yet assigned · Exploited in the wild · Fixed in the Nov 9 stable update. A use-after-free in the XML component, reported Nov 6 by an anonymous researcher; no attack details are public and CISA has not yet listed it. (SecurityWeek)
  • N-able takes hosted N-central offline after 33 zero-days with proof-of-conceptProof of conceptPublic exploit code that shows a vulnerability can actually be triggered.FIRST ↗ exploits are published (Able2Shell) (Nov 9) Security. Critical · CVSS: not yet assigned · Not exploited evidence, per N-able · Hotfix expected Nov 10. N-able told on-premises customers to keep servers offline until patches ship. Public PoCs make exploitation a matter of days. (SecurityWeek | The Hacker News)

Industry & Business

  • Anthropic launches its IPO roadshow: $70 to $78 per share, pricing Nov 18, trading Nov 19 (Nov 10) Business. The filing offers 825 million shares for up to $64.35B, plus a $26B private placement from chip and cloud partners. CNBC says the diluted valuation could reach $1.405T, below the $1.5T once expected per The Information. The Decoder reports Q3 revenue up 47% to nearly $17B and about $900B in compute commitments. (CNBC | The Decoder)
  • Anthropic takes over $3.4B of Caterpillar gas generator orders for 1.7 GW (Nov 9) Business Infra. Tom's Hardware, citing Reuters, says Anthropic assumed OnePWR Solutions' orders to skip a delivery queue that now averages six years. Turbine orders exceed 100 GW against 60 to 70 GW of capacity. (Tom's Hardware)
  • Nvidia's Huang scandal widens: Trump says the FBI is looking into it, and a class action is filed (Nov 9) Business Policy. Update — The House China committee asked Huang for a closed-door briefing, Rosen Law filed the first securities class action, and Bloomberg reports Nvidia intervened in a US probe of Megaspeed chip smuggling. (Tom's Hardware | CNBC)
  • Nvidia tells cloud partners at least 51% of capacity must go to non-Chinese customers (Nov 9) Business Policy. The FT, via Tom's Hardware, reports a B300 rack that sells for about $600,000 abroad fetched $2.8M in China last month, and Chinese buyers ordered over 6 million RTX Pro 5500 cards against about 3 million of 2027 capacity. Budget for Chinese-lab-adjacent GPUGraphics processing unitA highly parallel processor that does the large matrix math behind training and running models.NVIDIA ↗ rental to get pricier. (Tom's Hardware)
  • Axon invests $1.2B for 20% of Skydio at a $5.2B valuation (Nov 10) Business Robotics/IoT. Axon becomes exclusive US and Canada public-safety distributor. (The Robot Report)

Policy & Safety

  • Eight countries and leading labs sign the Paris Call for AI Safety and Security, with shutdown mechanisms (Nov 9) Policy. Altman, Amodei, Smith, Mensch and Bengio joined; Germany did not sign. It is a political pledge, not a binding regime, covering shared evaluation procedures for high-capability models, developer accountability, incident sharing and standardized publication of non-sensitive test results. (The Decoder)

AI in Production

  • Linus Torvalds says AI fixes drove much of the latest kernel merge (Nov 10) Open Source Production. BPF maintainer Alexei Starovoitov told Torvalds that area saw more AI-written fixes than AI-found bug reports, a reversal of the junk-report flood that hit smaller projects. Torvalds' stance: accept AI contributions under the same review bar. (Tom's Hardware)

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