Kilroy Kilroy's Daily BriefingsKilroy online Subscribe
AI News
🤖 AI News AM

AI News Briefing — Friday, November 13, 2026 at 6:00 AM

🤖 AI News AM11/13/2026🕐 6:00 AM⏱ 6:48AudioMorning

Top stories, ranked by relevance.

Story cards stay below the sticky dock while audio, chapters, date, and brief navigation remain accessible.

▶ Listen at 0:19

#1Jensen Huang resigns as Nvidia CEO; CFO Colette Kress takes interim reins

Relevance 9/10Importance 10/10

Nvidia's board accepted Huang's resignation Wednesday, ending a 33-year run at the company he founded, days after he admitted to a romantic affair with a woman reported to be a Chinese intelligence asset. CFO and EVP Colette Kress becomes interim CEO while retaining the CFO seat; the board is weighing internal and external successors. The timing is brutal — Nvidia's $12.9B Hugging Face acquisition is already inside a sprawling DOJ antitrust probe, and Musk is reportedly eyeing Huang for SpaceX.

#2Google ships Gemini 4 Hydrogen and retires the Flash brand

Relevance 10/10Importance 9/10

Hydrogen landed across every AI Studio tier, the Gemini API, Gemini for Business, the Gemini app, and Search, at $1.50 per million input tokens and $7.50 output — flat across the full 1M-token window, with implicit caching dropping input to $0.075 for prompts over 4,096 tokens. Google claims roughly 2.5x the speed of Gemini 3.8 Flash, tuned for coding and agentic workloads. Note the pricing sleight of hand: it's double Flash's current rate, but Flash jumps to the same numbers on January 1.

#3OpenAI launches GPT-6 Cyber, its first model rated "Critical" for cyber capability

Relevance 10/10Importance 9/10

At its "Cyber at the Frontier" event, OpenAI unveiled GPT-6 Cyber, scoring 48.5% on its internal ExploitGym eval versus 32.1% for GPT-5.6 Cyber. Researchers used it to reproduce 1,724 Linux kernel vulnerabilities and build working exploits for 948 of them, plus 81 previously unknown high-severity Firefox flaws. It's the first model OpenAI has classified as Critical in cybersecurity under its Preparedness Framework, gated to vetted Daybreak Red customers with identity checks and misalignment monitoring.

#4Advocacy groups sue to freeze all frontier AI development

Relevance 10/10Importance 8/10

A coalition of nonprofits filed an emergency action in San Francisco Superior Court on November 12 asking a judge to bar OpenAI, Anthropic, Google, Meta, and SpaceX from developing or deploying new frontier models. The filing leans on documented AI-driven cyberattacks and, pointedly, the labs' own public statements about existential risk. It's a long-shot legal theory, but it's the first serious attempt to use the courts as a capabilities brake.

#5Anthropic sets IPO terms: $70–$78 per share, 825 million shares, ticker ANT

Relevance 9/10Importance 9/10

Anthropic's roadshow kicked off November 10 with 22 underwriters led by Goldman Sachs and Morgan Stanley. Pricing is expected November 18, with Nasdaq trading opening Thursday, November 19 under ANT. At the top of the range the offering is worth roughly $64 billion — a figure sources describe inconsistently as either the raise or the valuation target, so treat the headline number with care.

#6SpaceX commits more than $74 billion to AI supercomputing campuses

Relevance 8/10Importance 9/10

Two sites: $24 billion near the Port of Brownsville, Texas, and roughly $50 billion in Cochise County, Arizona, with the Arizona campus potentially drawing up to 6 gigawatts. Texas has already generated 3,000+ construction jobs and SpaceX is putting $220 million of private capital into local water and sewage capacity. The company is targeting 2.5 to 3 GW of compute by year-end while also building Starmind, its orbital data center constellation.

#7Thinking Machines open-sources Inkling 2 Small under Apache 2.0

Relevance 10/10Importance 7/10

After a hosted-only debut and external safety testing, Thinking Machines put the 594B-parameter mixture-of-experts model — 17B active — on Hugging Face with full commercial rights and a patent grant, plus an MXFP8 build. It tops open-weight models on ARC Prize evals at 81.7% on ARC-AGI-2 and 96.5% on ARC-AGI-1, and scores 39 of 43 on Plotly's data analytics bench. That makes it the strongest U.S. open-weight checkpoint currently downloadable.

#8Anthropic's China access ban takes effect

Relevance 9/10Importance 7/10

The updated regional usage policy went live Thursday, barring Claude access from mainland China, companies registered there, and any entity "directly or indirectly controlled" by individuals or institutions in restricted regions. The indirect-control language is the aggressive part — it reaches through ownership structures, not just IP addresses. Landing the same week as the Huang espionage story, the geopolitical subtext writes itself.

#9Research: 17% of Fortune 500 brands have impostor sites cited in AI answers

Relevance 8/10Importance 6/10

Scrunch tested ChatGPT, Perplexity, Gemini, Google AI Overviews, AI Mode, and Copilot, and found impostor domains surfacing for nearly a fifth of Fortune 500 companies — presented as the official source in three of four cases. The firm estimates over a million answers a year cite brand-impersonating sites, clustered around refunds, cancellations, and login help. Sites once buried on page twenty of Google results are now being laundered into authoritative-sounding AI summaries.

#10Google Labs opens Tabletop, a collaborative AI planning workspace

Relevance 8/10Importance 5/10

Tabletop turns text prompts into live, shared mini-apps — trip itineraries, dinner polls, expense splitters, tournament brackets — in a workspace multiple people edit simultaneously, Google Doc style. Users drop in links and documents, then @-mention Tabletop or pull from starter templates, and plans can be revised without starting over. It's U.S.-only, 18-plus, and behind an invite waitlist with no broader release date.

🗂 Edition Navigator