Collection window: articles published between 2026-11-08 20:48 UTC and 2026-11-09 20:48 UTC (last 24 hours). All timestamps converted to UTC. Stories covered in the previous brief are excluded (for Nvidia, only the new board-probe development is included).
1. Nvidia’s board opens an independent probe into whether Jensen Huang was targeted by Chinese intelligence; Huang admits affair in staff memo
Nvidia’s board said on Sunday it is launching an independent investigation, led by independent directors with outside legal counsel, after reports that CEO Jensen Huang was the target of a Chinese spy operation. According to Bloomberg and others, Huang admitted in a company memo to an affair with Alice Huang, an executive at Singapore-based Megaspeed, writing that there is “no excuse for what I did.” The board said it has no reason to believe national security or company secrets were compromised and that it has confidence in Huang and his leadership team. China’s Foreign Ministry formally denied the spy-plot claim, and Huang faces a closed-door congressional briefing on November 18.
Reference: TechCrunch — 2026-11-08 21:00 UTC
2. Google adds TPU 8tx and TPU 8i Max, its first year with four TPU systems
Google SVP Amin Vahdat unveiled TPU 8tx and TPU 8i Max in a blog post. TPU 8tx pairs TPU 8t silicon with 5th Gen AMD EPYC hosts to deliver hardware-level confidential computing, aimed at private fine-tuning and secure agent execution for regulated industries and governments. TPU 8i Max packages two TPU 8i chips together, doubling compute, SRAM and HBM bandwidth per socket for long-context inference. Google also open-sourced TorchTPU for native PyTorch execution, launched the XLA Serving Engine and made Cloud Pathways generally available on GKE; both chips are slated for 2027.
Reference: Google Blog — 2026-11-09 · AI Weekly summary — 2026-11-09 13:57 UTC
3. Nvidia AI Day Seoul opens: SK Telecom plans a gigawatt-scale AI cloud, Coupang shows GB300 infrastructure
Nvidia AI Day Seoul 2026 opened at COEX. SK Telecom outlined plans for a gigawatt-scale AI cloud built on Nvidia’s DSX framework, with its first AI factory due in 2027, and Coupang Intelligent Cloud disclosed infrastructure built on GB300 NVL72 liquid-cooled racks. Samsung and SK Hynix presented PhysicsNeMo-based semiconductor device and process simulation, while LG Electronics and Doosan Robotics showed Isaac GR00T robot-training pipelines and a 2028 target for industrial humanoids. The event opened the same day Nvidia’s board announced its probe, drawing extra attention.
Reference: Tech Times — 2026-11-09 07:30 UTC
4. Trump on 60 Minutes: he “doesn’t mind” AI guardrails as long as the US stays ahead of China
In a post-midterm 60 Minutes interview, President Trump said AI risks are manageable and named the Department of Justice as the agency to handle them. He said he does not object to guardrails as long as the US does not fall into a “deep second position” behind China, and signaled openness to working with Democrats. He did not specify which safeguards he would support, and China came up eight times, keeping the focus on competition rather than safety.
Reference: CBS News transcript · AI Weekly summary — 2026-11-09 06:35 UTC
5. Meta takes Instagram’s AI age estimation global, moving suspected teens into Teen Accounts even with adult birthdays
Meta is expanding AI age estimation on Instagram worldwide, placing users it suspects are teens into Teen Account protections even if they list an adult birthday, The Verge reports. The classifier uses the profiles and content a user engages with, birthday messages from friends and visual cues in uploaded photos and videos; accounts judged to belong to under-13s are removed. Users can change settings if misclassified, but no error rate has been published. In Australia, Meta also uses Apple’s age-range signal, and Facebook and Threads are expected to follow.
Reference: The Verge · AI Weekly summary — 2026-11-09 12:36 UTC
6. AI-agent self-publishing platform Spines raises $64M as ARR climbs from $3.5M to $50M in two years
Israeli startup Spines is raising $64 million: $29 million in equity led by Zeev Ventures and $35 million in debt. Authors upload manuscripts and AI agents handle editing, proofreading, cover design and layout, with distribution to Amazon, Barnes & Noble and others in digital, audio and print-on-demand formats. ARR has grown from $3.5 million two years ago to $50 million, with 8,000 authors and more than 12,000 published books. Spines screens out AI-written books with detection tools and plans to expand AI translation to 20 languages within six months.
Reference: Globes — 2026-11-09 14:00 UTC
7. Unverified claim: ChatGPT Desktop forwards users’ SSH agent to remote hosts used by Codex
A Hacker News post claims that ChatGPT Desktop forwards a user’s SSH agent to remote hosts that Codex connects to, which could let a remote Codex agent push to private GitHub repositories using the user’s credentials. The only cited evidence, openai/codex PR #28131, was closed without being merged, which weakens the claim without disproving the behavior. Coming a week after reports that Claude Code’s “Don’t Ask Again” default leaked keys, it adds to scrutiny of how coding agents handle credentials. It remains a single-source claim pending confirmation from OpenAI.
Reference: Hacker News · AI Weekly summary — 2026-11-09 12:36 UTC
Today’s Summary
- Nvidia’s board probe shows the US-China security fight over AI chips has escalated into a CEO governance risk.
- Google’s four-TPU lineup and the big GPU builds from SK Telecom and Coupang show AI compute competition splitting into confidential computing, long-context inference and national-scale AI factories.
- Trump’s conditional openness to guardrails, Meta’s AI age estimation and the coding-agent credential debate show regulation and safety questions multiplying as AI spreads.
Leave a Reply