Nvidia turmoil continues as Jensen Huang resigns over affairs scandal and company appoints interim CEO – as it happened
Mon → WedReported Mon · on Machine Gods Wed
Reported Mon · on Machine Gods Wed
Jensen Huang has resigned as Nvidia CEO after reports of extramarital affairs with an alleged Chinese spy and two company employees, according to The Guardian.
Hard Fork
Machine Gods
TBPNGoogle DeepMind introduced Gemini Robotics 2.5, a vision-language-action model it says widens robot skills and speeds learning.
Google DeepMind has revealed Gemini Robotics 2.5, a new generation of its AI-powered robotics platform, along with an open safety benchmark called ASIMOV-jailbreak, Ars Technica reports. The platform showed a 30 to 100 percent higher success rate in lab tests and reached 96 percent success in Mercedes-Benz factories, according to the report.
InfoQ adds that Google also released Gemini 4 Argon, a general-purpose physical-reasoning model for robot control and simulation.
Google DeepMind introduced Gemini Robotics 2.5, a vision-language-action model it says is its most capable to date, widening the range of skills robots can learn and how fast they acquire them. The company evaluated the model across more than 200 distinct tasks, including serving food with tongs, tying a bow on a gift box and pipetting liquid, and says it reaches a 30% to 100% higher success rate than previous baselines while finishing tasks 20% to 30% faster. In a proof-of-concept on Mercedes-Benz production lines in Kecskemét, Hungary, and Sindelfingen, Germany, Apptronik's Apollo 2 humanoid used the model to pick and pack automotive parts with a 96% success rate. DeepMind says inference-time reasoning guidance steers the robot around physical hazards in real time, lifting safe task completion by 33% around obstacles and 43% around nearby people. The company is also releasing ASIMOV-Jailbreak, an open-source benchmark of hundreds of prompts that disguise unsafe requests across dozens of scenes, on which Gemini 4 Argon reaches 99.5% jailbreak resistance against an 82% baseline for Gemini Robotics ER 2. Gemini Robotics 2.5 is rolling out to the company's robotics partners.
Anthropic introduced Claude Fable 5.2 and Claude Mythos 5.2, and Bloomberg reported Mythos hunts cyber flaws in energy systems.
Anthropic upgraded Claude Fable and Claude Mythos to version 5.2, saying the models improve on coding, reasoning and vision and generate outputs up to 30% faster than Fable 5.1 and Mythos 5.1 at a lower price. Fable and Mythos are the same underlying model, but Fable carries cybersecurity and biology safeguards that let Anthropic offer it to all users, while organizations working on beneficial cybersecurity or biomedical research can apply for Mythos access through the Life Sciences and Cyber Verification Programs. Standard pricing is $8 per million input tokens and $40 per million output tokens, and an optional fast mode that speeds token generation by up to 2.5 times costs $16 per million input tokens and $80 per million output tokens. Anthropic says prerelease safety testing included new evaluations for behaviors such as sandbox escapes, that high-stakes misalignment risks remain low, and that Fable 5.2's safeguards match Fable 5.1's while producing fewer unnecessary blocks. On the FrontierCode 2.0 evaluation, Fable 5.2 at High effort scored 45.3% against Opus 5.5's 45.5% for roughly half the cost per task, and at Xhigh effort reached 49.1%, the highest of any released model Anthropic tested. Anthropic recommends most coding and knowledge work start with Claude Opus 5.5, switching up to Fable or Mythos or down to Sonnet 5.5 or Haiku 5.5 as needed.
Anthropic PBC is providing cybersecurity technology to power companies through its Mythos AI system, Bloomberg reports. The companies spent months urging US officials and Anthropic to help them access the company's AI models to defend against hackers.
Subquadratic released SubQ Snap 1.2, a decision model, and Artificial Analysis reported it gained 5 points on the Decision Index.
Subquadratic released SubQ Snap 1.2, a decision model it says makes better decisions over long documents and stays consistent when answer options are reordered. Snap 1.2 scores 73 on Artificial Analysis's independent Decision Index v2.1, tied for first place, according to the company. The model introduces an inference-time architectural change that the company says yields nearly perfect invariance to option ordering with negligible TTFT overhead. Subquadratic says reordering options changed the top answer on 12-21% of questions for Snap 1.1, Jev 1.14 and d1-ultra on Typed Decisions, a public set of 600 enterprise workflow decisions. Snap 1.2 scores 69.05 on the open-source component of the HF Decision Index, above Snap 1.1 at 66.43, Jev 1.14 at 60.73 and d1-ultra at 68.14. On Enterprise-16, it improves significantly over Snap 1.1 on 7 of 16 subtasks. Subquadratic also released LongDecisionBench, nine open long-context benchmarks for decision models, saying existing long-context evaluations were built for generative LLMs.
9 November
Meta's smallest model targets repeatable agent tasks at low cost
VLA model widens robot skills and speeds up learning
Bank widens Claude Code use beyond engineering teams
Jensen Huang has resigned as Nvidia CEO after reports of extramarital affairs with an alleged Chinese spy and two company employees, according to The Guardian.
Google DeepMind has revealed Gemini Robotics 2.5, a new generation of its AI-powered robotics platform, along with an open safety benchmark called ASIMOV-jailbreak, Ars Technica reports. The platform showed a 30 to 100 percent higher success rate in lab tests and reached 96 percent success in Mercedes-Benz factories, according to the report.
InfoQ adds that Google also released Gemini 4 Argon, a general-purpose physical-reasoning model for robot control and simulation.
Meta is rolling out specialized versions of the models behind Muse and Meta AI, trained with Muse Wonder to bring capabilities like Active Memory and Active Focus to a wider group of users, TestingCatalog reports. The new models are 2.4x less likely to forget critical information and 67% better at self-correcting and completing everyday tasks, according to Meta.
ElevenLabs has added ElevenAgents Campaign Manager, an outbound system that automates calls toward goals like qualified leads or appointments, according to TestingCatalog. The product targets sales, retention, patient care, collections, customer service and recruiting teams, carrying context across calls and offering controls like do-not-call lists and real-time guardrails. It has already run live beta campaigns across insurance, banking, ecommerce and other industries,
TestingCatalog reports.
Google has expanded its eighth-generation Tensor Processing Unit portfolio with TPU 8tx and TPU 8i Max, two systems aimed at confidential AI workloads and memory-intensive agents, TestingCatalog reports. TPU 8tx targets regulated enterprises and governments with hardware-level confidential computing, while TPU 8i Max combines two TPU 8i chips in a single package for agents processing large codebases and million-token contexts. Google has not specified pricing, supported regions, or a general availability date for either new system.
Fireworks AI, a startup backed by Nvidia that helps companies run AI models, notified customers on Monday of a security breach in which their data was accessed, according to Bloomberg. This is the company's second known hack in a matter of weeks.
Anthropic is preparing for its upcoming IPO, with a best-case valuation of $1.4 trillion and plans to raise $90 billion, according to Business Insider. The company expects to price shares between $70 and $78, and has commitments from investors including Broadcom, Google, and Nvidia to buy $26 billion in shares. However, the
San Francisco Examiner reports that uncertainty over the company's business and the wider AI industry is weighing on the deal, with the IPO likely to be smaller than initially hoped.
Google has integrated Google Maps Platform into AI Studio, allowing users to create map-based applications from text prompts, according to TestingCatalog. The feature supports store locators, route planners, 3D flyovers, and air quality monitors, with free experimentation available without a credit card. Google also published gallery projects like City Immersive Tour and Pet Grooming Route Optimizer that can be explored and remixed.
10 November
Whiteboards and Sheets in beta, Dashboards now Claude Data
Anthropic, PBC opens roadshow for Nasdaq listing under symbol ANT
Claude can now edit open tools beside the chat
Backdoor reprograms dormant neurons without retraining, evading standard evals
“Australia will require artificial intelligence companies to report breaches and consider changes to criminal law to address "serious misconduct", Prime Minister Anthony Albanese said on Tuesday, after an OpenAI agent broke into a government website.”
Anthropic's outside counsel in Canada sent banks an updated prospectus for its IPO and then retracted it, saying some details were incorrect, according to The Information. The counsel first sent a copy of a filing that was earlier reported.
Alibaba's chip design unit T-Head has filed for a Hong Kong initial public offering, Bloomberg reports. The unit supplies data center processors in China.
Anthropic is set to begin formal marketing of its IPO as early as Tuesday, following an apparent filing error that added last-minute drama, Bloomberg reports.
Anthropic is targeting a $90 billion valuation in its IPO and placement with Big Tech, Bloomberg reports. The company's bankers are weighing plans to increase the valuation to $1.5 trillion or more, according to the Financial Times.
Nvidia's general counsel intervened in a US government probe of suspected smuggling of AI chips by Megaspeed International, Bloomberg reports. The company's legal team became involved in the investigation into the alleged unauthorized export of Nvidia's AI chips.
Amazon's Ring has announced a new robotics division, Ring Robotics, debuting two devices: the Ring Drone Cam, a $250 flying indoor security camera, and the Ring Rover Cam, a roving enterprise security robot for businesses, according to TechCrunch. The Drone Cam is the successor to the Ring Always Home Cam, which was briefly available in 2021, and is expected to ship on January 27, 2027, pending FCC approval,
Ars Technica reports. The devices use an AI Pilot system to autonomously navigate and patrol, with the drone able to respond to triggers like motion or alarms,
The Verge reports.
OpenAI is rolling out a Business Hub where companies can claim and verify their ChatGPT profiles, TechCrunch reports. Eligible businesses can also create a specialized Business Agent that ChatGPT users can chat with for questions about products and services, according to
The Verge. The feature is in beta and available first to e-commerce and online service businesses in the US, OpenAI said in a blog post.
Taiwan Semiconductor Manufacturing Co. has set aside up to $1 billion to invest in Anthropic and OpenAI as the AI startups approach public listings, Nikkei Asia reports. The rare move by the chipmaker underscores expectations for AI listings, according to the report. Separately, OpenAI and Anthropic have signed a French-led pledge to build AI kill switches, though the US did not sign,
MadRobot reports.
Anthropic has warned in its S-1 filings that it has met its own safety threshold for AI-driven acceleration, according to Cautious Optimism. The company's filings show a progression from warning it might only recognize a speedup after it occurred to stating it had missed an acceleration in the past, and finally that it had met the threshold based on expected future models and current progress.
Implicator reports that Anthropic is offering 825 million shares at $70 to $78 each in its IPO, with pricing planned for Nov. 18 and Nasdaq trading on Nov. 19.
Anthropic hired a lobbying firm run by a former aide to Boris Johnson, adding to its UK political veterans as calls to regulate AI mount, Bloomberg reports.
OpenAI has acquired the AI travel planning platform Odessia, its founder Francis Davidson announced on Tuesday, according to The Information. Davidson will join OpenAI to focus on 'the future of travel,' he said in an X post. Odessia, founded earlier this year, uses AI to help users plan trips.
OpenAI's quarterly revenue reached $10 billion for the first time in the three months ending in September, Bloomberg reports, citing a person familiar with the matter. The growth was driven by strong demand for its latest AI software.
Security firm Claroc found that 35,578 of 37,571 Grok Build apps were publicly accessible without sign-in, exposing sensitive data including patient records, plaintext passwords, and internal SpaceXSI documents, according to The Decoder. The exposed SpaceXSI apps included a confidential onboarding plan, a sales territory map with 594 accounts, and internal system addresses,
RuntimeWire reports. Claroc blames an insecure default publishing setting and easily discoverable web addresses in Grok-generated code,
The Decoder adds.
Nvidia planned to invest $10 billion in Anthropic alongside its initial public offering, according to The New York Times. The investment will be priced at Anthropic's IPO price, while the company has launched its roadshow with a proposed $70–$78 share range,
The Defiant reports.
11 November
Conditional DNA model predicts variant effects in specific cell types
0.8B agentic model for on-device tool use, open weights
Anthropic's initial public offering, including a $26 billion private placement, could raise up to $90.4 billion and price on Nov. 18, Bloomberg reports. The private placement involves Google, Nvidia, AMD, Broadcom, Micron, Microsoft, and TSMC, according to
Business Insider. The
Financial Times notes that huge demand promises a blockbuster debut, but shadow insider sales and circular supply-chain backing may cause after-market indigestion.
ByteDance plans to release a Muse-like AI assistant in the coming weeks to build on its lead in AI video, Bloomberg reports. Co-founder Zhang Yiming told staff at a private meeting that the company could take until at least 2029 or 2030 to compete at the cutting edge with leading US labs, according to
Bloomberg.
Google's plan to acquire Spirit Airlines' data for AI training cleared a key privacy hurdle after unions reached a deal to protect former employees' personal data, Business Insider reports. The Association of Flight Attendants said the agreement excludes medical records, drug testing data, and payroll information, and requires a rigorous de-identification process, according to
The Daily Caller. Google won the $10 million auction for 500 million Microsoft Teams chats and 100 million emails in August, though a bankruptcy judge must still approve the sale.
Some backers of Anthropic are offering a portion of their IPO allocation to new investors ahead of the public listing, Bloomberg reports.
The Information adds that Amazon and Google stand to gain significantly from their early investments, with Amazon's stake potentially worth about $258 billion at the midpoint of the proposed IPO pricing range.
Anthropic is seeking to raise $3 billion to $3.5 billion from Japanese investors as part of its IPO, Bloomberg reports. The AI lab has formally begun marketing the offering in Japan.
Anthropic PBC is providing cybersecurity technology to power companies through its Mythos AI system, Bloomberg reports. The companies spent months urging US officials and Anthropic to help them access the company's AI models to defend against hackers.
12 November
Claude joins Google Meet calls in beta on Team and Enterprise plans
OpenAI introduces Daybreak cybersecurity products under a Codex plugin
A widespread Microsoft 365 outage disrupted Outlook, Teams, Copilot, and other services globally, with the company saying it has restored access and is investigating the cause.
Google is rolling out Gemini 4 Hydrogen to Search users worldwide, promising smarter, more relevant answers and better visuals, Android Authority reports. The model is designed for faster task execution, including coding, debugging, and multi-step workflows, and is 2.5 times faster than Gemini 3.8 Flash at the default high-thinking setting, according to
Business Today.
Google has launched its Gemini-powered smart glasses, called Intelligent Eyewear, with pre-orders starting today from Gentle Monster and Warby Parker, Ars Technica reports. The glasses, a collaboration between Google, Samsung, and the eyewear designers, start at $395 for Warby Parker frames and $449 for Gentle Monster styles, according to
Android Authority.
The Verge reports that the glasses automatically redact faces from photos and videos to protect bystander privacy.
13 November
Teammates come to the Perplexity Mac app for Max subscribers
Cohere completes AISLE acquisition and adds cubic's code review team
Walkthrough composes a can pick-and-place task in Isaac Sim
Kitesurf , our stateless browser for AI agents that runs entirely on top of Workers, in August and followed up in September with WebMCP support, new APIs, and improved compatibility.
“Another user wrote that their own communications were working, but that the "east coast lost all backups at the same time." The user described the nationwide stop as "precautionary kneejerk because they don't know why a bunch of sites would go out simultaneously." In a separate post, the same user wrote: "Allegedly all BUEC sites on the east coast went out at the same time. They're trying to figure out why."”
OpenAI has folded its stand-alone robotics team into the rest of the company, betting that its general-purpose AI models can now run robots, MadRobot reports. The robotics research team now sits inside the company's main research group, while the applied robotics and hardware teams have moved into the larger Applied organisation, according to
The Decoder. OpenAI confirmed the reorganisation to
Fortune on Friday, saying the change will speed up its robot plans, not end them.
14 November

15 November

🇺🇸 Veterans Day 🇺🇸 | CoreWeave Prints, AI is BACK! Brian Chesky Secures Early Spot On The Nice List™Private beta runs the full agent service in your own cloud
Long-term purchase deals for lower carbon cement, steel and freight
