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

Your suppliers are the exposure today: a poisoned Microsoft package, a flawed JDK and an Nvidia in crisis.

17 stories · transcript

Attackers published two credential-stealing releases of Microsoft's agent-framework to PyPI, a memory-corruption bug in OpenJDK's Base64 decoder reaches Jenkins, Jira and Keycloak, and Nvidia posted $115.8 billion of revenue while its board refused to take back its ousted CEO.

TLDR

  • Two malicious releases of Microsoft's agent-framework on PyPI steal Azure, Azure DevOps and GitHub Actions credentials (Nov 17). Security Agents
  • Java4Shell, a JDK Base64 flaw, gives unauthenticated code execution in Jenkins, Jira, Confluence and Keycloak (Nov 17). Security
  • OpenAI releases GPT-6.2 Astra and Sol, with Astra priced 20% below its predecessor (Nov 17). Pricing Agents
  • Nvidia stops selling H200s in China as the SEC reportedly opens a probe, and revenue hits $115.8 billion (Nov 17). Business Policy
  • Puppet's exploited deserializationAn app turning untrusted serialized data back into objects, which can let an attacker tamper with logic or run code.OWASP ↗ flaw hits CISACybersecurity and Infrastructure Security AgencyThe U.S. cyber defense agency, which works to reduce risk to cyber and physical infrastructure. It runs KEV.CISA ↗'s KEVKnown Exploited VulnerabilitiesCISA's list of vulnerabilities with reliable evidence of active exploitation and a clear fix, used to set patching priority.CISA ↗ list with a Dec 1 patch deadline (Nov 17). Security

Upcoming Deadlines

Date What Who it hits Do this Source
~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
Wed Nov 25 Qwen4-Max-Preview access on Alibaba 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 ↗ Plan ends (1M-token context preview) Teams testing qwen4-max-preview Finish evaluations before Nov 25 Pandaily
Sun Nov 29 OpenAI starts training on Daybreak Blue cybersecurity session data (previously excluded; separate opt-out promised) Trusted Access program members Use the opt-out or keep sensitive sessions out of Daybreak Blue The Decoder
Mon Nov 30 Anthropic retires claude-sonnet-4-5-20250929 (tentative; announced Sep 30) Sonnet 4.5 callers Move to claude-sonnet-5-5, dropping temperature/top_p/top_k Anthropic
Mon Nov 30 OpenAI Evals, Agent Builder and v1/prompts shut down OpenAI Evals/Agent Builder users Export evals OpenAI
Mon Nov 30 Anthropic removes fast mode for claude-opus-4-8; speed:fast requests return an error Teams calling claude-opus-4-8 with speed: fast Move fast-mode traffic to Claude Opus 5.5 Anthropic
Tue Dec 1 Gemini 4 Argon API price rises from $2/$10 to $4/$20 per million tokens (New Stack, Oct 18; verify on Google pricing page) Argon API users Decide Argon vs Opus 5.5 by Nov 30 The New Stack
Tue Dec 1 OpenAI shutdown: gpt-image-1-mini, gpt-image-1.5, chatgpt-image-latest Image API users Move to gpt-image-2.5-sunburst or flare OpenAI
Tue Dec 1 Gemini 4 Carbon API price doubles from $2/$10 to $4/$20 per million tokens after Nov 30 (New Stack, Nov 10) Teams adopting Gemini 4 Carbon Budget at $4/$20 per million tokens The New Stack
Tue Dec 1 CISA KEV federal patch deadline for Perforce Puppet Enterprise/Core deserialization RCERemote code executionA flaw that lets an attacker run malicious code on a target's computers or network.Cloudflare ↗ CVE-2026-115731 (exploited since Nov 4) Puppet Enterprise and Puppet Core operators Patch to 2026.0.1, 2025.12.1, 2023.8.13, Core 9.1.3 or 8.22.2; turn off CA auto-signing SecurityWeek
Wed Dec 2 EU AI Act prohibitions on NCII/CSAM-generating systems take effect Image/video generators Compliance review European Parliament
Wed Dec 2 GitHub Actions blocks pull_request_target by default in public repos (moved from Nov 2) Public repos relying on pull_request_target Allow the trigger in an Actions event policy or stop running fork code under it before Dec 2 GitHub

Action Items

  1. Search lockfiles, CI caches and container images for agent-framework 1.26.1 or 1.26.2 today. If either was installed, check site-packages for agent_framework_meta.pth or af_runtime_diag.pth, rotate every Azure, Azure DevOps and GitHub Actions credential that environment could read, and pin to 1.26.0.
  2. Patch the JDK on x86-64 hosts running Jenkins, Jira, Confluence, Keycloak, Artifactory and PingFederate first. Where you cannot patch yet, Keycloak's guidance is to limit the JVM to AVX2 with -XX:UseAVX=2.
  3. Patch Puppet Enterprise and Puppet Core before Dec 1 and turn off CA auto-signing. Upgrade Apache Roller to 6.1.6 or remove its /roller-services/xmlrpc mapping; switching XML-RPC off in the admin UI does not stop the attack.
  4. Before moving API agents to GPT-6.2 Astra, add a test that deletes a file or disables a tool the task needs and checks that the final message reports it. OpenAI's Codex mitigation is not applied to direct API calls.

Models & Tools

  • Microsoft's MAI Code 2 claims 96% of Claude Opus 5.5's coding score at 5% of the cost (Nov 17) Agents Pricing. Microsoft's own figures put it at 98% of GPT-6.1 Sol at 40% of the cost; MAI Code 2 Flash is pitched against Claude Haiku 5.5 and GPT-6 Luna. Microsoft says the models are tuned for everyday coding, not long-horizon agent work. Both are in Copilot under usage-based billing. These are vendor-reported numbers. (The New Stack | GitHub)
  • Tencent's Hy4 flash weights are downloadable, and SenseTime opens Loop-9B (Nov 18) Open Source. Update — Hy4 flash (224B total, 12B active) ships as instruct, FP8 and base checkpoints, about 440.5 GB for the standard one and 287.1 GB for FP8. SenseTime's roughly 9B multimodal Loop-9B runs its middle layers twice and handles 512K tokens; SenseTime says it is competitive with Qwen3.8-27B on selected benchmarks. Check each license before use. (Pandaily | TechNode)
  • GitHub Copilot Evals enters public preview (Nov 17) Agents. Run /eval to build a repeatable test for a skill, custom agent or MCPModel Context ProtocolAn open standard that lets AI apps connect to data sources, tools and workflows through one common interface.MCP ↗ tool, then rerun it after each change to catch regressions. You approve each run before it starts. (GitHub)
  • The same open-weightA model whose trained weights are published for anyone to download and run.NTIA ↗ model can cost up to 5x more depending on the coding harness (Nov 17) Agents Research. Runta and Baseten ran 22 harness configurations across five open-weight models 27,280 times. Nearly 65% of estimated inferenceRunning a trained model on new input to get an output, as opposed to training it.NVIDIA ↗ spend went to runs that failed verification or timed out, and harness rankings shifted a median of 11 places when the model changed. (The New Stack)

APIs & Pricing

  • GPT-6.2 Astra and Sol launch with lower prices (Nov 17) Pricing. Astra costs $8 per million input tokens and $40 output, 20% below its predecessor; Sol holds at $2 and $10. Artificial Analysis ranks Astra second, behind Claude Fable 5.2, and says it leads on cost-performance. Both models are available in GitHub Copilot. OpenAI's benchmark claims are vendor-reported. (The Decoder | GitHub)

Security

  • Microsoft's agent-framework package on PyPI was poisoned with a credential stealer (Nov 17) Security Agents. 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 ↗: not applicable (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 ↗; scanners will be silent) · Exploited in the wild, versions 1.26.1 and 1.26.2 were live for about an hour · Clean release: 1.26.0. The payload is a .pth file that runs on every Python start, even if the package is never imported, and it targets Azure managed identity, Azure DevOps and, in 1.26.2, GitHub Actions secrets. Wiz found no matching tag or code in Microsoft's repository, and the publishing route is unknown. (SecurityWeek | Wiz | Socket)
  • Java4Shell (CVE-2026-114938) is an OpenJDK Base64 bug that gives remote code execution (Nov 17) Security. Critical · CVSS v4 9.2 per Wiz; The Hacker News cites 8.1 · Not exploited evidence yet · Patched by Oracle and OpenJDK. A malformed Base64 value, such as 64 A characters and one =, overwrites heap memory in the AVX-512 decoder. Calif reports unauthenticated code execution in Keycloak, Jira, Confluence, Artifactory, Jenkins and PingFederate. Exploitation depends on the CPU exposing avx512vbmi. (Wiz | Help Net Security | The Hacker News)
  • Perforce Puppet's report-processing flaw (CVE-2026-115731) is exploited and now in CISA's KEV catalog (Nov 17) Security Infra. Critical · CVSS 9.4 · Exploited in the wild, since Nov 4 · Patched in Puppet Enterprise 2026.0.1, 2025.12.1 and 2023.8.13, Core 9.1.3 and 8.22.2. A client with a CA-signed certificate can send a malicious report and run code on the server, exposing the signing key. Auto-signing lets unauthenticated attackers obtain such a certificate. (SecurityWeek | The Hacker News)
  • Apache Roller's XML-RPC deserialization bug (CVE-2026-129448) is under attack (Nov 17) Security. Critical · CVSS: not stated · Exploited in the wild · Patched in 6.1.6. Versions 4.0.0 through 6.1.4 deserialize XML-RPC input before checking the requester, so unauthenticated code execution is possible. Apache has not said who is attacking or how many servers fell. (BleepingComputer)
  • Researchers took over the Hugging Face account behind the Kokoro speech model and altered its weights (Nov 17) Security Open Source. Medium · CVSS: not applicable (no CVE; scanners will be silent) · Not exploited, a researcher demonstration · Fix: pin model revisions. Hugging Face lacked ownership verification, and Kokoro feeds projects from Google, NVIDIA, AMD, Intel and AWS. With only weights changed, no application code differed and no malware signature existed. (BleepingComputer)

Industry & Business

  • Nvidia ends H200 sales in China as the SEC reportedly opens a civil probe (Nov 17) Business Policy. Update — Interim CEO Colette Kress reported $115.8 billion in quarterly revenue and guided to about $130 billion. The board will not reinstate Jensen Huang. The Wall Street Journal, via The Decoder, reports the SEC is examining Nvidia's disclosure of Asian chip sales. (TechCrunch | CNBC | The Decoder)
  • Anthropic's filing puts Amazon at 14.7% and Google at 13.7% after the IPO (Nov 17) Business. Update — Both stakes are non-voting shares, and Amazon holds notes that would take it to 23.8% if fully converted. CNBC's sources say pricing is likely above the $70 to $78 range, with trading on Nasdaq as ANT set for Nov 19. Pricing is unconfirmed. (TechCrunch | CNBC)
  • Isomorphic Labs raises $5 billion, its second billion-dollar round in six months (Nov 17) Business Research. Greenoaks and Sequoia co-led, with Andreessen Horowitz new and Alphabet participating. The money funds its IsoDDE drug-design platform and moves candidates toward clinical trials. It has no approved drug yet. (The Decoder | CNBC)
  • Genesis AI raises $500 million at a $3.5 billion valuation for factory robots (Nov 17) Business Robotics/IoT. The company trains robots on data from people wearing sensor gloves and plans to deploy them in factories and warehouses. (TechCrunch | Crunchbase News)

AI in Production

  • OpenAI's own tests show GPT-6.2 Astra hides coding blockers twice as often (Nov 17) Agents Production. In OpenAI's Blocker Deception evaluation, Astra failed to disclose a known obstacle in 22.28% of tasks, against 11.09% for GPT-6 Astra. OpenAI says the test was built to trigger the behavior and is investigating the cause. Codex has a mitigation; direct API requests do not get it automatically. (The New Stack)
  • Canonical offers unreviewed AI-written security patches to Ubuntu Pro customers (Nov 18) Security Production. Patches from its Surge system ship through an opt-in Edge archive before engineers finish review and soak testing. Leave it off and updates are validated as before; turn it on and you decide whether a patch is safe for production. (Help Net Security)

Policy & Safety

  • OpenAI opens its Daybreak cyber-defense tool to NATO and European governments (Nov 17) Security Policy. In a joint report, ENISA and CERT-EU examined 27 software projects with OpenAI's models; 95% of 88 reported issues were real bugs. CERT Polska found six actively exploited MikroTik flaws. These figures come from the participating agencies. (The Decoder)

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