AI News
2026 year · Nov 9onwardPublic sources · Weekly overviewPublished every Monday

AI News Weekly

AI developments selected from public sources

Issue 609Nov 2026Sunday
2026 year6 issues
20
major stories
66
sources
9
first-party stories
About 9 min read

Issue guide

This issue compiles 20 important events from 7 published daily briefs.

Lead story

Huang Renxun Involved in Extramarital Affair with Chinese Spy, Nvidia Board Launches Investigation

This issue compiles 20 major events from 7 published daily reports.

Public sourcesBBC ChineseAlso 7 other sources reported thisFollow-upOriginal ↗
01

Model releases / updates

AWS Launches Claude Fable 5.2 and Claude Mythos 5.2

AWS announced that Claude Fable 5.2 and Claude Mythos 5.2 are now available on Amazon Bedrock and Claude Platform on AWS; the former targets complex tasks such as coding, mathematics, and scientific research, while the latter is available only through Anthropic's verification program for cybersecurity and life sciences scenarios. Fable 5.2 is listed as a Covered Model, requiring by default up to 30 days of data retention and human safety review, and eligible customers can use it with zero data retention until June 30, 2027. Claude Code's claude-api skill can assist with migrating model IDs, thinking settings, and cache configurations via the /claude-api migrate command.

Original

Google DeepMind Releases Gemini 4 Hydrogen, Targeting Everyday Coding and Agent Workflows

Google DeepMind released Gemini 4 Hydrogen, focused on speed and efficiency for everyday coding and agent workflows; on the Artificial Analysis Intelligence Index it scores 51 at the default high thinking level, above Argon's 50, and can reach 54 at the xhigh level. The model's 1 million token context window is priced uniformly at $1.50 per million input tokens and $7.50 per million output tokens, and it offers four thinking levels: low, medium, high, and xhigh. Google also launched Orchestrator in Antigravity, which can dynamically switch models between Carbon and Hydrogen within the same session, approaching Carbon's task success rate on Terminal-Bench 4.0 while consuming on average 24% less quota and running 19% faster.

Original

Meta Releases Lightweight Model Muse Ember, Focused on Cost-Effectiveness for Agent Tasks

Meta has released Muse Ember, its smallest and most efficient model, aimed at complex reasoning execution and everyday agent work; it can carry workflows across applications, call tools in sequence, run tests, and make targeted code fixes. The model supports a 1M-token context window and natively understands text, images, audio, and video. On the Artificial Analysis Intelligence Index, its performance is comparable to Claude Haiku 5.5, at roughly one-third the per-task cost, and it scores higher than GPT-6 Luna under every reasoning setting. Muse Ember is now live on the Meta Model API, priced at $0.10 per million input tokens and $0.50 per million output tokens.

Original

Baidu releases ERNIE 5.5 Flash Preview, focused on end-to-end programming, and announces upcoming open source

Baidu has released ERNIE 5.5 Flash Preview, with 280 billion total parameters and 14 billion activated per inference, support for a million-token context, a generation speed exceeding 200 tokens per second, and an announcement that it will soon be open-sourced. The model focuses on strengthening programming and agent capabilities, and the company says it can compete with Claude Opus 5.5 and GPT-6.1 Sol on benchmarks such as Terminal-Bench and AutomationBench. The author tested it by orchestrating animations with Python code, generating a playable "Backrooms" web game, and building a Suzhou garden in Blender. The weights are expected to be released soon under Apache 2.0 with support for SGLang and vLLM deployment; the API is already live on Baidu AI Cloud's Qianfan platform, with a limited-time 50% discount of 1 yuan input and 4 yuan output per million tokens.

Original

Claude Fable 5.2 launches on Claude Platform and Claude Code

Claude Fable 5.2 is now live on Claude Platform and Claude Code, up to 30% faster than Fable 5.1, priced at $8/$40 per 1M tokens, down from the previous $10/$50. effort defaults to high on Claude Platform and medium in the Claude app, with xhigh reserved for the hardest long-running agent tasks. Fable 5.2 also offers a Fast mode with 2.5x faster output at double the price, set via "speed": "fast" in the API, and Claude Code can switch to it in the model selector.

Original
02

Product release/Updated

Google Releases Gemini Decisions API and Gemini Decision 1 Model

Google launched the Gemini Decisions API, powered by the new model Gemini Decision 1 (gemini-decision-1), which can evaluate up to 1,000 yes/no, multiple-choice, and rating questions on text, image, audio, video, and PDF inputs in a single forward pass, returning confidence scores from 0 to 1. The model has an end-to-end latency of 150-250 ms and is priced at $0.10 per million input tokens, with output tokens not billed; an API key is available in Google AI Studio.

Original

Google Releases TPU 8tx and TPU 8i Max, Expanding Eighth-Generation TPU Product Line

Google expanded its eighth-generation TPU product line with the addition of TPU 8tx for confidential inference and TPU 8i Max for ultra-long context and the highest per-rack compute density, forming four systems together with TPU 8t and TPU 8i; this is also the first time Google has launched four TPUs in a single year. TPU 8tx is based on the same chip as TPU 8t and paired with fifth-generation AMD EPYC CPU hosts, protecting data and model weights in use at the hardware level, with up to 2x lower latency than TPU 8t; TPU 8i Max packages two TPU 8i chips together, with twice the compute, SRAM, and HBM bandwidth of TPU 8i. Google also open-sourced TorchTPU, TPU Sync, and the XLA Serving Engine, and announced the general availability of Cloud Pathways and GKE dynamic slicing.

Original

AWS releases AWS Emulator, a local emulator written by autonomous AI agents

On November 11, 2026, AWS released AWS Emulator, a local emulator built on 438 AWS API models and covering more than 200 services, allowing developers and coding agents to call AWS APIs on their own machines without an AWS account. Matthew Goldberg, an AWS Senior Principal Engineer, said the team neither wrote nor reviewed the emulator's code; humans handled specifications, the testing mechanism, merge gates, and the release pipeline, while three agent teams were responsible for writing tests, locating problems, and fixing them, fixing thousands of issues per day. The code is released under CC0 (vendor code under Apache 2.0), the repository does not accept PRs, and defects can be reported via GitHub Issues; the emulator does not evaluate IAM policies and is not suitable for production workloads or security testing.

Original

OpenAI launches ChatGPT Wallet, allowing dot agents to make payments on users' behalf

OpenAI has released ChatGPT Wallet, which lets dot agents complete purchases on users' behalf, in partnership with American Express, MasterCard, and Visa. Before each purchase, users must verify the item, amount, and selected card, then confirm via a security check such as Face ID or a fingerprint; the payment network then provides a payment token that substitutes for the real card number, so neither the merchant nor the agent ever receives the actual card information. The feature is initially available to a limited group of ChatGPT Pro users and cannot be used on websites that block agents.

Original

Anthropic releases Claude Sheets, Whiteboards, and Data, three AI office tools

Anthropic has released three AI office tools—Claude Sheets, Claude Whiteboards, and Claude Data—with the new features rolling out on web, desktop, and mobile starting that day. Claude Sheets supports generating and editing spreadsheets with formulas and tabs together with Claude, Claude Whiteboards can turn brainstorming content into documents or PPTs, and Claude Data offers authenticated queries and real-time sharing, with authenticated queries and real-time sharing still in testing. Anthropic also upgraded interactive connectors, allowing partners to embed applications such as Adobe and Atlassian directly into Claude.

Original

DeepSeek opens App capabilities such as search, files, and voice to the API

DeepSeek has opened the search, file input, URL reading, safety checks, and voice capabilities behind the DeepSeek App to the API; V4.1-Flash (deepseek-flash) supports these built-in capabilities in Messages, Responses, and Chat Completions, with tool calls and context assembly handled server-side. Developers can enable them via parameters such as deepseek_web_search and deepseek_read_url, or call standalone APIs such as OCR, ASR, and TTS separately, and an open-source front-end demo that supports importing DeepSeek conversation history is provided. Extended capabilities are billed separately based on tool usage on top of standard token fees; public-facing applications must enable safety checks, and content moderation, model registration, and AI-generated content labeling remain the developer's responsibility.

Original
03

Industry news

Anthropic Launches Roadshow for Proposed IPO

Anthropic announced that it has launched the roadshow for its proposed initial public offering (IPO). The company also stated that the registration statement for the relevant securities has been filed with the SEC but has not yet become effective, and that prior to effectiveness, securities may not be sold nor offers to buy accepted, and that this post does not constitute an offer to sell or a solicitation of an offer to buy.

Original

Huang Renxun Resigns as Nvidia CEO, CFO Colette Kress Appointed Interim CEO

Nvidia announced that Huang Renxun has resigned from the positions of president, CEO, and director, effective immediately; current Executive Vice President and CFO Colette Kress will serve as interim CEO while continuing in her original role until a permanent successor is appointed. After the announcement, Nvidia's pre-market share price once fell to $150.13, down about 13.58% from the previous trading day's close; after the open the decline remained above 10%, narrowing to -4.63% before publication. The board has launched a formal CEO search that will consider both internal and external candidates, and appointed Harvey C. Jones as the company's first board chairman in its history; the company confirmed it will report third-quarter fiscal 2027 results on November 17 as originally scheduled.

Original

Nvidia CEO Huang Renxun Resigns, CFO Colette Kress Appointed Interim CEO

Nvidia's board announced that Huang Renxun has resigned from the positions of president, CEO, and board member, effective immediately, and that Chief Financial Officer Colette Kress will serve as interim CEO; the board said it will consider both internal and external candidates. The resignation comes four days after the company announced an internal investigation on November 8 concerning his relationship with Alice Huang; people familiar with the matter told WIRED that the relationship with a subordinate and business dealings with Megaspeed may have had a greater influence on the board's decision than the Chinese spy allegations. Huang Renxun co-founded Nvidia in 1993; in the summer of 2025 the company became the first ever to surpass a $4 trillion market capitalization, and as of September 2026 it held about 90% of the GPU market.

Original

Anthropic Reportedly Plans to List on Nasdaq Next Week, Valuation Expected in the $1.1 Trillion to $1.75 Trillion Range

According to people familiar with the matter, Anthropic plans to list on Nasdaq next Thursday or Friday under the ticker ANT; this week it will hold roadshow meetings with core investors and is expected to file documents early this week disclosing the IPO price range. Market expectations for its valuation range from $1.1 trillion to $1.75 trillion, corresponding to an expected fundraising size of $80 billion to $100 billion. Foreign media analysis noted that if market conditions deteriorate further, the company may still postpone the IPO.

Original

Nvidia CEO Jensen Huang admits to extramarital affair; board launches investigation

On November 8, Nvidia CEO Jensen Huang admitted to employees that he had an extramarital affair with Alice Huang, the head of Singapore-based customer Megaspeed, whom U.S. intelligence suspects was recruited by China's Ministry of State Security to get close to him. Nvidia's board has commissioned an external firm to conduct an independent investigation, saying it has no reason to believe company secrets or national security were compromised, but that it still trusts Huang. After the news broke, Nvidia shares fell 4% to 6% in premarket trading and are down 18.60% over the past month, leaving its market value behind Apple's.

Original

Anthropic withdraws IPO pricing document; draft had sought to raise up to $69.3 billion

Reuters reported that Anthropic sent banks a draft securities offering on the evening of November 9 seeking to raise up to $69.3 billion through an IPO, followed by a follow-up email saying the pricing details were inaccurate, with the company saying it would update the document. The draft showed 866.7 million Class A shares priced at $70 to $80, of which Anthropic was selling about 843.3 million and existing shareholders 23.3 million; there was also a $26 billion private placement of Class N common stock involving Google, Nvidia, AMD, Broadcom, Micron, Microsoft, and TSMC. The report said Anthropic's quarterly revenue for the quarter ended September 30, 2026, was close to $17 billion, with quarterly growth having slowed markedly.

Original

Anthropic opens IPO subscriptions, seeking to raise up to about $90.4 billion

Anthropic plans to issue 825 million shares at $70 to $78 per share, alongside a $26 billion private placement, for a combined maximum raise of about $90.4 billion, which could surpass SpaceX's planned IPO to become the largest stock offering in the world. The company is seeking a fully diluted valuation of about $1.4 trillion, with Goldman Sachs and Morgan Stanley leading underwriting and 20 other banks participating, and plans to list on the Nasdaq Global Select Market under the ticker ANT. Some existing investors are reselling IPO subscription allocations or selling indirect interests through SPVs, raising market concerns about selling pressure in the early days of trading.

Original
04

Research papers

Google DeepMind Formally Verifies Signal Protocol Code with AlphaProof Code

Google DeepMind researchers used the AlphaProof Code agent framework to complete formal verification of key security properties of the actual Signal protocol code, generating more than six million lines of Lean proofs over the course of four weeks. The framework divides work among multiple LLM agents, with human experts determining the security goals to be proven and reviewing about one percent of the code. The team also reintroduced two previously fixed real vulnerabilities into an isolated version of the Signal code, and the system produced machine-checked insecurity proofs for both. DeepMind also open-sourced the formal verification artifacts.

Original
End of issue

AI News WeeklyAutomaticallysummarizedfrom public reports, with source links · View archive