The AI trade now depends on one person's judgment and on borrowed money, and both were questioned in the same weekend.
Nvidia's board opened an independent investigation into CEO Jensen Huang's tie to a suspected Chinese intelligence asset linked to chip customer Megaspeed, while CoreWeave's bonds yield above 16% going into its earnings. Underneath, builders got concrete tools: Google's new TPUTensor Processing UnitGoogle's custom chip built to speed up machine learning, especially large matrix operations.Google Cloud ↗ systems, a PyPI provenance tab and a patch for a live OpenSSH bypass. It is a lighter Monday: 16 items, because the weekend news cycle was dominated by one story.
TLDR
- Nvidia's board opens an independent investigation into CEO Jensen Huang's relationship with a woman US intelligence calls a suspected Chinese informant tied to chip customer Megaspeed (Nov 8). Business Policy
- Attackers are exploiting the OpenSSH "SSHole" authentication bypass, CVE-2026-123158, on musl-based Linux such as Alpine; fixed in 10.6p2 (Nov 9). Security
- Google adds TPU 8tx for confidential inferenceRunning a trained model on new input to get an output, as opposed to training it.NVIDIA ↗ and TPU 8i Max with double the memory per rack, its fourth TPU system this year (Nov 9). Infra
- Acronis says one Taiwan-based attacker ran 12 AI agents, mostly on abliterated open models, against emergency services in at least 11 countries (Nov 9). Security Agents
- CoreWeave's 2031 bonds yield above 16% as neocloud credit risk overshadows its Tuesday earnings (Nov 9). Business Infra
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 | 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 | |
| ~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
- Upgrade OpenSSH to 10.6p2 on every Alpine, Buildroot, OpenWrt and Yocto host today, or set
PasswordAuthentication noin sshd_config and restart sshd; then search auth logs for logins that used passwords of 257 to 1,024 characters. - Inventory every public repo with a
pull_request_targetworkflow before the Dec 2 default block: today's 16:00-18:00 UTC brownout is the first dry run, with another on Nov 16. - Turn on PyPI's new Security tab check for your top Python dependencies, and flag any release that lost attestations or changed repository or workflow since the previous version.
- Check any signing service that uses PKCS#1 v1.5 with 1024-bit RSA keys (HSMs behind PKCS#11 included) and schedule migration to 2048-bit or larger keys this quarter.
- Re-run your agent-misuse threat model with local, safety-stripped models in it: block outbound scanning from developer machines and log default-credential logins on internal admin panels.
Security
- Attackers are exploiting the OpenSSH "SSHole" bypass (Nov 9) 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 ↗: not yet assigned in the source · Exploited in the wild (botnets and crypto miners so far) · Patched in OpenSSH 10.6p2. CVE-2026-123158 lets a password of 257 to 1,024 characters log in to accounts with no password hash, such as root, when sshd runs on musl libc without PAM: Alpine, Buildroot, OpenWrt, Yocto. Previdian saw 20 attempts from 13 IPs in 9 countries between Nov 6 and Nov 8. Update — exploitation is now confirmed. (SecurityWeek)
- A lone attacker ran 12 AI agents against emergency services in at least 11 countries (Nov 9) Security Agents High · CVSS: 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 ↗; default credentials and SQL injection) · Exploited in the wild · Patch: none. Acronis watched the BOSSAI platform for 12 days. Commercial models mostly refused, so agents ran on abliterated DeepSeek, GLM and Qwen variants, including Cursor pointed at one. They pulled Taipei dispatch records and 206,800 Mexican phone records. A scanner invented CVEs and over half its high findings were false positives. (The Hacker News)
- DeepMind agents cut the cost of attacking 1024-bit RSA (Nov 9) Security Research Low · CVSS: not applicable · Not exploited · Patch: use 2048-bit or larger keys. Agents reduced a PKCS#1 v1.5 forgery to 391 CPU core-years from 1,380, in under four days with about 20 human messages. Key recovery was not achieved. The theoretical many-key factoring speedup of about 30x has not been implemented. Matters only for legacy HSM-signing setups. (SecurityWeek)
- PyPI adds a Security tab showing release attestations (Nov 9) Security Open Source Low · CVSS: not applicable · Exploited in the wild: Not applicable · Patch: not applicable. The tab marks releases with full, partial or no provenance, and warns when a release comes from a different repository or workflow than the last one, or when attestations disappear. Missing provenance is not proof of malice; treat a change as a prompt to check maintainer history. (Help Net Security)
Models & Tools
- Cursor adds realtime voice calls with its agents (Nov 6) Agents Voice chat runs on Grok Voice in the Agents Window, IDE, web and iOS. The voice agent hands coding work to the chat agent, narrates progress, and answers questions about diffs and runs while work continues. Privacy Mode gives zero data retention. (Cursor)
- Perplexity's Teammates keeps agent state in Git, not in the session (Nov 8) Agents Each project is a repository, agents work in separate sandboxes and submit changes back, and memory lives in its own repositories. Perplexity says under 2% of submissions are rejected for stale state, with median latency under 1.7 seconds (company-reported). Worth copying for any multi-agent setup. (The New Stack)
- Simon Willison: put agent-written PR descriptions inside
<details>(Nov 9) Agents Write a short human-approved summary on top and collapse the accurate but exhausting agent text beneath it. A cheap fix for review fatigue. (Simon Willison)
Infra
- Google ships TPU 8tx and TPU 8i Max (Nov 9) Infra The 8tx keeps data and weights encrypted during processing for regulated workloads; the 8i Max puts two 8i chips in one package for double compute, memory and bandwidth in the same rack. Both reuse existing silicon. Google says TPUs have nearly 100 customers and supports PyTorch through open-source TorchTPU, vLLM and SGLang. (Google Cloud | The Decoder)
- Google open-sources TPU Sync and takes Cloud Pathways to GA (Nov 9) Infra TPU Sync moves KV cache across HBM, host memory, SSD and Lustre; Google reports over 90% lower KV transfer latency and TTFT for 235B-480B MoEMixture of expertsA model made of many sub-networks, called experts, where a learned router sends each input to only a few of them.Shazeer et al. 2017 ↗ models (vendor-reported). Pathways brings single-controller JAX across TPU slices. (Google Cloud)
- SemiAnalysis: Envision built 120MW of a planned 800MW AI campus in under a year (Nov 9) Infra The Inner Mongolia site, at one-fifth the cost of comparable US projects by SemiAnalysis's estimate, pairs with its own wind farms and storage. It targets roughly Abilene-scale capacity by end of 2027. (Tom's Hardware)
- GitHub Actions and webhooks degraded on Nov 9 (Nov 9) Infra Run start delays and webhook delivery failures were mitigated and resolved; GitHub promises a root cause analysis. Separate anonymous-traffic timeouts also hit that morning. Check for missed webhook-triggered deploys. (GitHub Status)
Industry & Business
- Nvidia opens a board investigation of Jensen Huang (Nov 8) Business Policy Huang told staff he had an affair with Megaspeed executive Alice Huang, whom intelligence agencies reportedly call a Chinese informant. The board says it has no reason to believe secrets were compromised. Shares fell about 5% premarket, and incoming Democratic committee chairs promise oversight, which weakens Huang's lobbying against tighter export controls. (CNBC | The Decoder)
- CoreWeave's credit signals flash red before earnings (Nov 9) Business Infra Its 2031 bonds yield above 16% and five-year credit default swaps sit above 1,240 basis points. CoreWeave reports Tuesday and Nebius Thursday. IREN says 48% of backlog is Microsoft. If you rent neocloud capacity, ask about customer concentration and contract terms. (CNBC)
- Chinese labs recruit in San Francisco (Nov 9) Hiring MiniMax lists about 16 San Francisco openings, including a core-model research lead, and Moonshot is hiring US solutions architects. Rest of World says poaching top US-lab talent will be hard. (Rest of World)
AI in Production
- Pinterest swaps journey clustering for a fine-tunedTraining an already trained model further on extra data so it fits a specific task, domain or style.Anthropic ↗ 4B Qwen3 model (Nov 9) Production Customer AI The model reads up to 150 actions over 360 days and generates ranked journey names. Pinterest reports email click-through up 1.1% and push opens up 1.3% (company-reported), a modest gain from collapsing a multistage pipeline into one small model. (InfoQ)
Policy & Safety
- IEEE Spectrum: location verification for AI chips detects diversion, but only afterward (Nov 9) Policy Ping-based checks time round trips to trusted servers; they miss offline chips and can be spoofed with stolen keys, so physical audits remain the follow-up. A House committee bill would require verification, and the Nvidia probe raises its odds. (IEEE Spectrum)