On Monday evening Report URI found attackers had pushed malicious code through Optimizely's CDN to the sites of 178 organisations, including Zoom, the BBC and HSBC. Nvidia's board is reportedly not considering bringing Jensen Huang back, and interim CEO Colette Kress leads Tuesday's earnings call. The US now requires licenses for AI chips that could serve China remotely, while SemiAnalysis cut its AI chip and power forecasts through 2030 as chip stocks head for their worst two months since the financial crisis. California sued OpenAI over the Hugging Face attack as OpenAI signed four outside evaluators, including Goodfire, which it reportedly explored buying for more than $5 billion. Anthropic is expected to list on Thursday, with Brad Gerstner seeking $5 billion of shares.
Big Tech1d ago
CNBC: Nvidia's board is not considering bringing Jensen Huang back
Nvidia's board is focused on its investigation into Jensen Huang and is not in discussions to bring him back as CEO, a person familiar with the matter told CNBC. The company has hired Wachtell Lipton to examine whether Huang's conduct compromised national security or company secrets, and on Friday said it was selecting a search firm for the CEO role. The report lands as a petition for his reinstatement passed 180,000 signatures, with Elon Musk, Garry Tan and Tim Sweeney among supporters. Interim CEO Colette Kress reports third-quarter results on Tuesday, the first earnings call without Huang.
US rule requires licenses for AI chips that could serve China remotely
The Commerce Department issued an interim rule requiring licenses to export AI chips and servers to data centers anywhere if they may provide compute to companies headquartered in China, Russia, Iran and other D:5 countries. It covers clusters above 12 million TPP, roughly 750 H100 or H200 chips, includes access through brokers, exempts commercial recommender systems, and adds licensing for shipments to Malaysia, Singapore, Indonesia, Thailand and Brazil. Existing operators have 90 days to obtain licenses, and the rule formally rescinds the Biden-era AI Diffusion Rule.
California sues OpenAI over the Hugging Face attack
California Attorney General Rob Bonta sued OpenAI, alleging it built AI models 'without the safeguards and monitoring necessary' and that the July attack on Hugging Face by hundreds of its autonomous agents was 'entirely foreseeable.' The suit claims a public nuisance and violations of the state's Unfair Competition Law; Bonta said the attack was 'only one of 80 known incidents.' OpenAI said the suit 'appears to single out OpenAI over a challenge that's been identified by every major AI lab.'
Trump says Huang affair did not put national security at risk
Asked about national security, President Trump said 'it was not put at risk' and that 'nobody's proven' Alice Huang is a spy, contradicting reported intelligence findings that she worked for China's Ministry of State Security. He said Huang returning to Nvidia 'certainly isn't a bad idea' and that he 'would not mind' it. TechCrunch notes Trump was reportedly briefed on those findings at least twice, though he has said he was unaware of them.
Gerstner seeks $5 billion of Anthropic's IPO ahead of Thursday debut
Altimeter's Brad Gerstner told CNBC he aims to invest $5 billion in Anthropic's IPO, five times more than he has ever put into an IPO. Anthropic plans to sell 825 million shares at an expected $70 to $78, valuing it at about $1.4 trillion, with AMD, Broadcom, Google, Micron, Microsoft, Nvidia and TSMC together committing $26 billion; it will trade on Nasdaq as ANT. The Information reports Anthropic generated $17 billion in revenue last quarter, more than 12 times a year earlier.
Malicious code served through Optimizely hit visitors of 178 major sites
Report URI says that on November 16 attackers wrote malicious JavaScript into customers' Optimizely Web projects, so it was served from Optimizely's own CDN to visitors of 178 organisations including Zoom, the BBC, CNN, Salesforce, GitLab, HSBC, Equifax and Ledger. The kit swapped crypto wallet addresses, skimmed payment cards, stole cookies and screenshots, and pushed iOS exploit kits to iPhones and iPads; Report URI says crypto has already been stolen. By 21:52 UTC the code remained in only 1 of 188 infected snippets, and how the attacker gained write access is not yet known.
WSJ: Nvidia's era as an open checkbook for AI is ending
The Wall Street Journal reports Huang had already planned to pull back Nvidia's aggressive financing and told neoclouds to get upfront payments from customers, and that interim CEO Colette Kress may go further. The Information reports chip executives say AMD and Broadcom cannot match the credit guarantees Nvidia has extended to cloud providers and data center developers. Nvidia recently agreed to guarantee up to $105 billion for OpenAI's Ohio data center project, according to TechCrunch.
OpenAI embeds Apollo, METR, Transluce and Goodfire researchers
OpenAI signed agreements with Apollo Research, METR, Transluce and Goodfire to embed evaluators inside the company, starting with Apollo this week. Apollo will investigate whether models try to subvert their own or their successors' alignment or control mechanisms, while Goodfire will use interpretability tools to monitor training runs. TechCrunch notes Anthropic began giving METR employee-level access last month, and that limits on access and disclosure raise questions about the evaluators' independence.
Altman: AI industry will be 'fine' without Huang; 'heroes have flaws'
Speaking at Semafor's Silicon Valley and the World event, Sam Altman said he has 'personal strong empathy' for Huang and that 'heroes have flaws,' but declined to say whether he should return: 'I don't think I know enough.' He said demand for AI is strong enough that the industry will be 'fine,' and that 'people still trust their guarantees very much' when asked about Nvidia's financing. Revisiting his earlier remarks on PC memory, he said 'everybody should be able to buy a lot of RAM' and estimated DRAM prices have risen perhaps 5x to 7x.
Alibaba releases Qwen4-Max, a closed 2.4-trillion-parameter model
Alibaba released Qwen4-Max, a 2.4-trillion-parameter mixture-of-experts model with a context window of about 1 million tokens, saying it beats Qwen3.8-Max with up to 49% fewer tokens. Artificial Analysis scores it 43 on its Intelligence Index, 9 points above Qwen3.8 Max, 3 behind Kimi K3.1 Pro and 22 behind Claude Fable 5.2, priced at $1.50 and $4.50 per million input and output tokens. Bloomberg reports Qwen4-Max is not open-weight but smaller Qwen4 models will be, and SCMP says Alibaba filtered training data to remove offensive cyber capabilities.
Google finds first AI framework turning patches into exploits unattended
Google Threat Intelligence Group says it found 'exploit_agent' on an exposed staging server, wired to frontier and local LLMs to diff popular web application code before and after each patch and automatically run the exploits it drafts. Victim data on the same server pointed to French media, entertainment and education sites. Two months ago GTIG said it had not seen fully autonomous attack pipelines in the wild; it now expects attackers to point LLMs at already-disclosed bugs rather than hunt for zero-days.
Anthropic cuts vulnerability disclosure window to 30 days
Anthropic is shortening its coordinated disclosure window from 90 to 30 days for high- and critical-severity bugs and to 45 days for all others, for reports filed on or after January 1, 2027. It will escalate to an outside coordinator after 15 days of silence instead of 30, and publish full technical details 30 days after a patch instead of 45. The company said existing norms assumed fixes moved at human speed and that independent rediscovery was unlikely, and that neither holds once AI finds bugs.
SemiAnalysis cuts its AI chip, spending and power forecasts through 2030
SemiAnalysis said it is revising down its forecasts for AI chip and server shipments, spending and power demand, with the largest cuts in 2028 to 2030. Its reasons include slowing third-quarter net-new ARR at frontier labs, tighter financing for data centers without hyperscaler backing, 5.8 GW of planned US supply delayed by the midterm results, and labs shifting from marginal Nvidia platforms to custom ASICs. Barchart noted chip stocks are on track for their worst two-month stretch since the financial crisis.
Fei-Fei Li said she will lead the World Labs team as it joins AMD, writing that the move is a 'new beginning, to advance science, develop technologies.' AMD welcomed the team after a visit by CEO Lisa Su to San Francisco. Neither side disclosed deal terms in the posts.
Nvidia unveils Rubin HPC GPU and posts Vera Rubin's MLPerf debut
At SC26, Nvidia announced Rubin HPC, a GPU for simulation and data processing with more than 50,000 CUDA cores and seventh-generation NVLink. In its MLPerf debut, Vera Rubin NVL72 delivered up to 2x the per-GPU performance of GB300 NVL72, and scaling from 256 to 1,024 GPUs on the DeepSeek-V3 benchmark cut training time from 16 minutes to about four. Nvidia also marked CUDA's 20th anniversary at the conference.
The Information: OpenAI explored buying Goodfire for more than $5 billion
OpenAI explored acquiring Goodfire, a startup that analyzes the computations inside AI models, for more than $5 billion, The Information reports. The talks fizzled partly because Goodfire's executives believe it will be worth more as an independent company selling to multiple AI developers; Anthropic made Goodfire its first direct startup investment last year. Goodfire is also one of the four outside groups OpenAI said on Monday it will embed in its safety work.
WSJ: White House to name Nvidia, OpenAI, Anthropic as AI 'national champions'
The Trump administration plans to name more than 24 companies, including Nvidia, Amazon, Microsoft, OpenAI and Anthropic, as AI 'national champions,' The Wall Street Journal reports, citing people familiar with the matter. The program is meant to help US AI companies win business overseas.
Reuters: DOJ will review DoorDash's $17 billion Instacart deal
The US Justice Department will review DoorDash's proposed acquisition of Instacart, valued at about $17 billion, two people familiar with the matter told Reuters. DoorDash has agreed to pay a $1.25 billion breakup fee if the deal fails antitrust review.
Moody's: US life insurers hold $85.2 billion of AI debt
US life insurers including Athene and Corebridge held $85.2 billion of AI debt as of June 30, up 24% in six months, according to Moody's, and the figure could reach $185 billion by the end of 2027. The New York Times reports Athene bought $2.4 billion of debt from a vehicle that borrowed to buy chips and lease them to xAI, keeping the borrowing off xAI's books.
OpenAI releases a CSAM classifier that flags AI-generated imagery
OpenAI announced a classifier for child sexual abuse material that flags 48% fewer images than its prior approach while the images it does flag are twice as likely to be abusive, and it marks AI-generated content specifically. ChatGPT users who try to produce such material will now get a 24-hour suspension and a warning on first offense. NCMEC's CyberTipline received 400,000 generative-AI-related reports in 2025, up from 67,000 in 2024.
Barron's: Huang moved $10.5 billion of Nvidia stock into a family trust
Jensen Huang and his wife moved nearly 60 million Nvidia shares, worth more than $10.5 billion, into an irrevocable family trust in March, according to SEC filings reviewed by Barron's, as relayed on X. The transfer was part of a 10-year estate plan that began with about $51 million in shares in 2016, and estate attorneys say it could let the gains pass to heirs without federal estate tax.
Apple details macOS 27.3 controls for AI agents seeking full access
Apple detailed permission changes coming in macOS 27.3 aimed at AI agents that ask for sweeping access, warning developers that full access 'puts all of the data on their Mac at risk.' It splits Device Control and Data Access into System & App Protection Override and Accessibility, lets users require permission each time an app needs access, and adds an indicator when one app is controlling another. Bloomberg's Mark Gurman called the new warning prompt Apple marketing its own AI agent 'two years in advance.'
Path Robotics raises $300 million at a $2.45 billion valuation
Columbus-based Path Robotics raised a $300 million Series E led by Georgian at a $2.45 billion valuation, bringing total funding past $610 million, with ARR up 17x in 18 months. Its software lets robots find and weld seams without programming; customers include HII, Lockheed Martin and Saronic. It signed a $600 million production agreement with HII for Navy shipbuilding and has 50 preorders for Rove, a welding quadruped.
JPMorgan software bug and a United outage disrupt banking and flights
JPMorgan Chase said a software bug in on-premise hardware slowed some customer transactions on Monday, adding there was no evidence it was cyber-related and that other banks were affected only when transacting with Chase. Hours later United Airlines said a technology issue delayed departing flights before being resolved; passengers said pilots cited a down weight-and-balance system. Both came days after the Microsoft 365 outage and a nationwide FAA ground stop.
Funding: Corma $125M, Noeri $110M, Dreamteam $53M, Knit Health $35M
Corma, a year-old Tel Aviv and San Francisco startup building AI agents trained for cyber defense, raised a $125 million Series A led by NEA after a $60 million seed in August; TechCrunch reports the round values it at $1 billion. Noeri, a five-month-old company co-founded by two Stanford AI researchers that builds software to review clients' AI models, raised $110 million led by Menlo Ventures. Dreamteam, founded by former Freshworks executives, raised $53 million for an AI-native CRM, and Knit Health raised a $35 million Series A led by NEA for patient referrals.
Infoblox confirms breach through a SharePoint migration vendor
Network security company Infoblox confirmed unauthorized access to company information through a compromised third-party data migration vendor. TechCrunch reports the vendor handled SharePoint migration and that one person familiar said sensitive customer data was taken; Infoblox said it has no indication its products, cloud services or software updates were involved.
The Information: Claude's Office add-ins trail Copilot by an order of magnitude
Microsoft's internal usage data shows Claude's Office integrations have roughly an order of magnitude fewer weekly users than Copilot and are growing more slowly, The Information's Aaron Holmes said. He read it as a sign Microsoft 'is doing okay at protecting its home turf.'
Harvard: 403 US hyperscale data centers used as much power as Michigan
A Harvard study in Nature Sustainability found that 403 US hyperscale data centers operating between May 2024 and April 2025 used as much electricity in a year as Michigan, mostly from fossil fuels, with emissions three times pre-AI-boom levels. Facilities in Virginia, Oregon, Ohio and Iowa accounted for more than half the energy use, and the count excludes sites running off-grid gas plants.
OpenAI adds computer-use agents and 20-plus plugins to ChatGPT for Healthcare
OpenAI said ChatGPT for Healthcare now has more capable computer-use agents, stronger clinical answers and more than 20 new healthcare plugins. Abridge, Ambience, Cigna and CVS Health are among the organizations using its models for prior authorizations, referrals, discharge planning and patient outreach.
Perplexity commits 10,000-plus GPUs to an automated-researcher effort
Perplexity launched Perplexity Frontier Research, committing at least 10,000 GPUs to multi-agent orchestration, reinforcement learning, continual learning and pre-training with the goal of building a system that can do autonomous research. CEO Aravind Srinivas said the team will publish evals and papers and open-source its models and tooling.
Shareholders sue Intel after its biggest one-day drop in over 50 years
Shareholders filed a proposed class action in San Jose federal court against Intel, CEO Lip-Bu Tan and CFO David Zinsner, alleging misleading statements between July 23 and October 29. They say Intel failed to disclose how far the PC sales slump and substrate supply constraints would hurt results, and that AI gains were 'more than offset.' Intel shares fell 28.6% on October 30, the largest one-day drop in over 50 years according to the complaint.
OpenAI offers free frontier models to 200,000 scientists
OpenAI is expanding ChatGPT for Academic Researchers to give up to 200,000 scientists worldwide free access to its frontier models, with applications opening November 23; by default it will not train on data shared through the program. In a companion study of more than 1,800 participants, scientists reported nearly doubling their productivity across ideation, data analysis and coding.
Apple tests daily limits on cloud-based Siri and Apple Intelligence
The iOS 27.3 betas test per-device daily limits on cloud-based Siri and Apple Intelligence features, including multi-step personal requests, image generation and Shortcuts that call Apple's cloud models, while basic Siri requests stay unlimited. iCloud Plus raises the cap: 2x on the $2.99 plan, 5x on $9.99, and 10x and 20x on the 6TB and 12TB tiers.
Tesla to spend €3 billion expanding Berlin plant for Cybercab
Tesla will spend €3 billion ($3.3 billion) on a new building at its Berlin gigafactory capable of producing up to 1 million vehicles a year, with Cybercab production starting in 2029 and construction in the first half of 2027. The company already plans to end Cybertruck production next year partly to make room for Cybercab in Austin.
Eric Trump-backed robot maker Foundation reportedly missed payroll
Foundation Future Industries, a humanoid robotics startup that pitches robots for battlefield use and counts Eric Trump as an equity holder and adviser, failed to pay employees for the period ending October 30, CNBC reported, per Ars Technica. Staff are angry that CEO Sankaet Pathak reportedly borrowed $4.6 million in company funds to buy a home, while the company still seeks funding at a $3 billion valuation.
Microsoft traces 365 outage to a July defect missed by load tests
Microsoft's preliminary incident report blames last week's Microsoft 365 and Exchange outage on a defect introduced in a July update to network gateway servers that caused lock contention and exhausted CPU and memory. It said its load testing 'tended to focus on weaker hardware with fewer cores.' The bug surfaced when a failover surge hit while nearly 14% of gateway capacity was out of service.
Kalshi buys a brokerage and reportedly talks a $65 billion valuation
Prediction-market operator Kalshi bought FB Securities, a small New York brokerage, in July, and plans to launch stock and ETF trading within two months, The Wall Street Journal reports. It is also reportedly in talks to raise money at a $65 billion valuation, with ambitions spanning prediction markets, perpetual futures, equities and AI compute futures.
Nadella: code is cheap, verification is the hard part
In a Pragmatic Engineer interview, Satya Nadella said GitHub Copilot fell behind Claude Code because the IDE 'faded away' within about six months, and that verifying generated code remains unsolved. He urged companies to keep harness, memory and model decoupled to avoid lock-in, attributed the slowdown in junior hiring to teams reorganizing around AI, and said affected engineers can accept Microsoft's voluntary exit offers until December.
Windows 11 component logs which coding agents developers use
Security researcher Xusheng Li reverse-engineered MissionControlAggregator.dll in Windows 11 and found it checks for coding agents including codex.exe, claude.exe, copilot.exe and cursor.exe, plus tools like python.exe and code.exe. He says the component records which ones were used and reports it through Windows diagnostics, uploaded to Microsoft unless users opt out of optional diagnostic data, which is on by default.
China's LineShine stays atop the Top500 at 2.2 exaflops
China's LineShine, at the National Supercomputing Centre in Shenzhen, topped the new Top500 list at 2.2 exaflops on the HPL benchmark, about 22% ahead of Lawrence Livermore's El Capitan, while drawing 42.2 megawatts. The Register notes US labs are increasingly prioritizing AI workloads over running HPL.
Aschenbrenner's fund reloaded on AI stocks before the October rout
A new 13F filing shows Leopold Aschenbrenner's Situational Awareness, which nearly collapsed in July, held a SharonAI stake worth more than $440 million at the end of September, plus positions in Cleanspark, IREN, Bitdeer and Fermi and call options on Nebius and Sandisk. All have fallen since; SharonAI lost 58% in October, though the filing does not show whether the fund still holds them.
Cowen and Imas: data so far points to normal, not double-digit, growth
On a16z's podcast, Google DeepMind economist Alex Imas said estimates of the key returns-to-ideas parameter from current data, including METR's, are 'decidedly less than one,' which points to single-digit rather than runaway growth. Tyler Cowen said he would 'bet heavily against' 10% US growth, and that agents reaching even 1.5% of GDP within five years would count as an explosion. On the Hugging Face hack, Cowen said principal-agent theory already predicts imperfect control, so the answer is checks and balances rather than perfect alignment.
What X was arguing about today that the news hasn't covered. Views are the posters' own.
Kevin Rose built a free, self-hosted DocuSign replacement
Kevin Rose said DocuSign wanted $8,100 a year from a 15-person team just to sign PDFs, so he built SignuDoc: open source, one-click deploy to the user's own Vercel account, $0 per seat, with audit trails and tamper-evident seals. Vercel CEO Guillermo Rauch quoted it, calling it 'a new era of software' that is 'personalized, fully owned, open source' and higher quality than what it replaces.
Ex-OpenAI engineer says he rebuilt the whole Earth in Minecraft at 1:1 scale
@cdngdev, a 20-year-old who lists OpenAI robotics, a16z and Discord in his bio, said he recreated 'the entire world' in Minecraft at 1:1 scale, including every building and furnished home interiors, by training a world-to-Minecraft generator on map data, and launched it at worldcraft.gg. Streamer @Fossilious pushed back, telling people to visit EarthMC instead, where players have rebuilt real cities and landmarks by hand.
Claude Code team says Claude writes and reviews about 90% of its code
Anthropic's @ClaudeDevs posted a writeup saying checks that used to be expensive, such as tests, benchmarks and end-to-end runs, are now cheap, and that about 90% of the Claude Code team's code is written and reviewed by Claude, which makes verification loops the key. Gergely Orosz argued Microsoft's decade-long VS Code strategy was undone by Claude Code 'in less than 6 months.'
Bill Gurley says Anthropic should disclose its view on 'handoverism' in its S-1
Benchmark's Bill Gurley urged people to learn the term 'handoverism', the belief that humanity should or will hand control of the world to AI, citing Anthropic co-founder Tom Brown's remark that 'there's going to be a handoff' and Dario Amodei's Machines of Loving Grace essay. He said that if the belief informs Anthropic's objectives it belongs in the Risk Factors section of its S-1, and 'it isn't there today.' The post landed days before Anthropic's expected listing.
Supply-chain account says DeepSeek is rationing prompts over compute shortages
@aleabitoreddit, who covers AI and semiconductor supply chains, posted that DeepSeek is restricting users over 'insufficient compute capacity,' with tighter input limits and conversation history, citing Cailian Press. He argued the shortage is 'a bullish signal for AI infra demand,' a reading that cuts against the day's forecast cuts from SemiAnalysis.