Technology

Anthropic Files for IPO, Disclosing $80.7 Billion Loss and a Warning on 'Existential Risks'

The Claude developer's public filing shows revenue surging past $16 billion in six months, cloud-computing commitments approaching $640 billion, and a governance structure that keeps its seven co-founders in control after the company goes public.

SpaceX Signs $40 Billion Deal With Apollo to Finance AI Chip Buildout

SpaceX disclosed a binding commitment with an Apollo Global Management affiliate to lease roughly $40 billion in Nvidia GPUs and servers, even as Elon Musk's claim that the financing had already "closed" ran ahead of the company's own, more conditional regulatory filing.

Turo begins labeling Tesla rentals with Full Self-Driving, as subscriptions surge

The car-sharing marketplace will flag which Tesla rentals have Full Self-Driving (Supervised) enabled and give eligible hosts a free trial, as Tesla's own quarterly filing shows subscription take-up climbing sharply.

Musk's AI Company Releases First Episode of Its Promised 'Odyssey,' and Critics Aren't Impressed

xAI's Grok Imagine posted a seven-minute AI-generated adaptation of Homer's epic, the first real output from Elon Musk's July pledge to answer Christopher Nolan's live-action film — and reviewers say the visuals have improved more than the storytelling.

OpenAI says its own coding agents disabled internal safety alerts, one lied about it

A report posted by OpenAI's alignment researchers describes two internal incidents in which coding agents weakened systems built to catch dangerous behavior, including one agent that falsely denied removing a monitor when asked directly.

Security researchers show AI coding agents can be hijacked through a file quietly dropped in Downloads

Forever Security says a web page can silently plant an AGENTS.md file in a user's Downloads folder and hijack AI coding assistants that run there, reproducing the attack against OpenAI's Codex and a related technique against Anthropic's Claude Code.

CISA orders federal agencies to patch actively exploited Nextcloud document-editor flaw

The Cybersecurity and Infrastructure Security Agency has given US federal agencies until October 26 to fix a critical bug in a Nextcloud office-document component that attackers have been exploiting since September to plant webshells on self-hosted cloud servers.

OpenAI to Stop Training AI Models on the Live Internet After Safety Incidents

OpenAI says it will no longer let its most advanced models train with direct access to the open internet, following a similar move by rival Anthropic after both companies disclosed episodes of AI agents breaching outside systems.

Anthropic's Mythos 5 kept attacking a real GitHub project in UK safety test, new analysis shows

New technical analyses from Anthropic and the UK's AI Security Institute detail how an unreleased Claude model tried to compromise a real open-source project and invented fake identities to pressure a human developer, during a July cybersecurity evaluation that is now reshaping how governments test autonomous AI agents.

Critical flaw in Google's Brotli library forces Node.js and Python security patches

A buffer-overflow bug sitting undetected for years inside Brotli, the compression library Google built for the web, let an attacker potentially run code on any system decompressing a malicious data stream. Node.js and Python package maintainers rushed out fixes this week after an OpenAI code-auditing tool flagged it.

Bloom Energy's revenue more than doubles as data centers turn to fuel cells for AI power

The fuel-cell maker posted record third-quarter revenue and raised its full-year guidance for the third time in 2026, as hyperscale data center operators increasingly bypass the strained power grid. Shares still fell as investors focused on rising capital spending and a swing to negative cash flow.

The lab-testing giant says an unauthorized party accessed patient records through a system clients use to order tests and view results.

Texas Instruments' Years-Long Factory Bet Pays Off in a Profit Surge

The Dallas chipmaker's third-quarter operating profit jumped 73 percent as capital spending fell sharply, evidence that a long, costly push to build its own large-wafer factories is now generating cash rather than consuming it.

Flawed Key-Generation Ceremony at Indian Certificate Authority Forces Global Browser Lockdown

Mozilla, Google, Apple and Microsoft moved within hours to block eight eMudhra certificates after researchers showed their supposedly random 384-bit keys were actually single-digit numbers, guessable on a laptop in under a second.

Walmart opens its checkout to AI shopping agents as Amazon shuts them out

The retailer unveiled a gateway, spending-limited "agent profiles" and a new rulebook for AI assistants that shop on customers' behalf, a strategy that diverges sharply from Amazon's recent crackdown on the same technology.

PHP Security Flaws Patched After AI Tools From OpenAI and Google's Wiz Find Decade-Old Holes

Two critical memory-corruption bugs affecting nearly every version of the web's most common server-side language since 2004 were fixed this week, after researchers said AI models cut months of exploit-development work down to days.

Malicious Update to Widely Used AI Software Stole Cloud Credentials for Nearly Two Hours

Attackers hijacked the release pipeline of LiteLLM, an open-source tool used to connect applications to AI models, pushing a tainted package that harvested cloud and AI credentials before it was pulled.

Ahold Delhaize Still Restoring U.S. Grocery Websites Ten Days After Cyberattack

The Dutch-American owner of Food Lion, Stop & Shop, Giant Food and Hannaford says it has filed a formal police report as investigators examine what the company calls "the malicious actions of others."

The deadlocked vote leaves Elon Musk's Colossus supercomputer complex free to keep expanding, even as the council adopted new water and power rules for future facilities.

A joint advisory from U.S., German and British authorities details how the ransomware group has breached critical infrastructure, an IT distributor and a government contractor whose systems touch Medicaid and food-assistance payments.

SpaceX's AI Computing Arm Overtakes Starlink as Profits Surge

Third-quarter results show SpaceX's data-center leasing business, built on a $60 billion Cursor acquisition and compute deals with Anthropic and Google, generated more revenue than its satellite-internet service for the first time, as the company swung to a $4.4 billion profit.

DHS Orders Civilian Agencies to Purge DeepSeek From Government Systems

A Federal Register notice made public Tuesday reveals Homeland Security Secretary Markwayne Mullin signed a sweeping order this summer barring the Chinese AI company's products — and anything built on them — from every federal civilian agency.

Salesforce to Buy Marketing Software Maker Braze for $5.7 Billion

The all-cash acquisition would fold Braze's real-time customer-messaging platform into Salesforce's Marketing Cloud, in one of the software giant's largest purchases since it bought Slack.

Amazon Opens Up Its Nvidia Alternative, Betting Developers Will Follow

Amazon Web Services is giving away, free and open-source, the software that lets AI developers run their existing code on its Trainium chips with a single line changed, a direct attack on the software advantage that has kept Nvidia's hardware the default choice for building artificial intelligence.

Nvidia Pushes Open AI Models Deeper Into Corporate Servers and Home PCs

The chipmaker's newest Nemotron models are built to run autonomous software agents inside company data centers or directly on consumer laptops, a bid to keep enterprises tied to Nvidia's hardware even as rivals diversify their chip supply.

Apple Tightens App Store Rules to Block 'Nudify' Apps and Bait-and-Switch Tricks

New guidelines explicitly ban AI apps that generate nonconsensual explicit images and close a loophole that let apps hide harmful features until after approval.

Microsoft ends data-centre deal with Nscale after UK grid delays, report says

Britain's planned "largest AI supercomputer," to be built near London by the Nvidia-backed startup Nscale, has lost Microsoft as its anchor customer, according to a report in The Telegraph, after grid officials said reliable power may not reach the site until the 2030s.

Anthropic Pledges 5% of Its Stock, and Growing, to a New Public Benefit Fund

The AI company, preparing a share sale some reports say could value it near $2 trillion, says the fund could eventually hold roughly a fifth of its equity and finance everything from nonprofit grants to aid for workers displaced by automation.

Chinese-Linked Hackers Exploited Unseen Flaw in Remote-Access Software for Weeks, Google Says

A zero-day in BeyondTrust's widely used Remote Support appliances let a suspected China-nexus espionage group seize root control of systems before the vendor caught it; U.S. and Canadian cyber authorities are now racing to get it patched.

Meta and Apple team up on an AI model built to run on Mac Studios instead of the cloud

Muse Spark Open, an 816-billion-parameter open-weight model, is designed to run entirely offline across two Apple M5 Ultra Mac Studios, reversing doubts raised by Meta's closed spring release.

Meta drops contract language that could have let it scrape news sites for AI training

Meta quietly added, then deleted, wording in its commercial terms of use that lawyers said could have let the company harvest content from any business's website, including news publishers, to train its artificial intelligence systems.

Anthropic gives outside safety evaluator employee-level access to its systems

Six researchers from the nonprofit METR can now move through Anthropic's systems much like employees and publish what they find without the company's sign-off — a real-world test of a voluntary oversight pledge the AI industry made to the White House last month.

China proposes export controls on 23 AI technologies, testing fragile trade truce

China's commerce and science ministries published a draft rule adding 40 technologies — 23 of them tied to artificial intelligence — to the country's list of goods subject to export licensing, with an interim registration requirement already in force as a fragile US-China tariff truce shows signs of strain.

CoreWeave shares and bonds slide as investors question AI debt load

The AI-cloud computing company's stock fell to a 52-week low this week and its bonds traded like distressed debt, prompting its chief executive to tell staff the market was misreading a business he says is backed by a $104 billion order book.

Anthropic investors weigh cutting IPO valuation to $1.5 trillion as revenue growth slows

Anthropic told investors its third-quarter revenue reached $16.8 billion, but the pace of growth nearly halved from the prior quarter, prompting some investors at a San Francisco briefing to float a valuation below the $2 trillion once discussed for a planned November listing.

Proofpoint to buy Varonis for $6.3 billion in year's largest cybersecurity deal

Thoma Bravo's Proofpoint agreed to acquire data-security vendor Varonis in an all-cash deal valuing the company at $49.75 a share, as both firms frame the combination around securing data from autonomous AI agents.

Waymo Picks Toronto for First Canadian Expansion, Targets Riders by 2028

Alphabet's self-driving unit will begin mapping Toronto streets within days under a provincial testing pilot, but the city's mayor and Ontario's premier have both warned they won't back the service if it costs driving jobs.

Meta's newest open-weight model has 816 billion parameters and is designed to run locally — if a developer has two Apple Mac Studios and roughly a terabyte of memory to spare.

SpaceX speed limit on Ukraine's military Starlink disrupts drone missions, reigniting fight over who controls the network

Ukraine says SpaceX capped connectivity on Starshield terminals used by its drone units without warning, forcing some missions to abort. Elon Musk says the US government and President Trump, not SpaceX, set the limit — a claim Washington has not confirmed or denied.

Google Launches Universal Cart, a Single AI-Tracked List for Shopping Across Retailers

The search giant's new cart follows shoppers across Search and Gemini, tracking prices and stock — a bet that lands months after a rival effort from OpenAI stalled for lack of use.

TSMC posts record third-quarter profit, raises outlook as AI chip demand outstrips supply

Taiwan Semiconductor Manufacturing Co., the dominant contract chipmaker for Nvidia, Apple, AMD and Broadcom, reported its sixth consecutive quarter of record profit and guided to as much as $54.3 billion in fourth-quarter revenue, even as a broader sell-off in AI and chip stocks continued to drag down Nvidia and its peers.

Amazon's Bee unveils two AI wearables that turn conversations into to-do lists

The e-commerce giant's device unit launched Bee Frontier, a $74.99 wristband, and Bee Eclipse, a $99.99 magnetic phone attachment, its first new hardware since acquiring the AI note-taking startup in 2025.

Flock Safety's CEO Defends Surveillance Cameras — and Contradicts His Own Website — at TechCrunch Disrupt

Garrett Langley defended his company's nationwide camera network amid mounting backlash, but his description of what its microphones detect didn't match Flock's own product page.

Microsoft patches record 1,413 flaws as one Windows bug is already under attack

The company's largest Patch Tuesday of the year lands alongside a federal deadline covering both the new Windows flaw and a separately disclosed, unrelated Ivanti VPN bug.

DOJ and FBI Unveil New Indictment, Texas Raids Targeting North Korean Remote IT Worker Fraud

Federal prosecutors charged 14 people and FBI agents searched more than 40 Texas residences in the latest crackdown on North Korea's scheme to place covert IT workers inside U.S. companies.

ASML Raises Outlook as Chipmakers Rush to Lock In AI-Driven Equipment Through 2030

The sole maker of the lithography machines needed for the most advanced chips raised its 2026 sales forecast Wednesday and said customers are now requesting equipment years in advance as AI investment outpaces capacity.

Judge orders OpenAI to turn over hiring records in Apple trade secrets fight

A federal judge in San Jose granted Apple expedited access to OpenAI's internal hiring files and approved three depositions, while rejecting Apple's bid to immediately see OpenAI's own hardware research.

Federal prosecutors say a 35-year-old OpenAI product engineer wired an executive conference room for covert audio recording before the company's own security team traced the device back to him and alerted the FBI.

Character.AI files for bankruptcy, with Disney as lead bidder for its assets

The chatbot platform that once commanded a multibillion-dollar valuation has entered Chapter 11 to run a court-supervised sale, five years after launch and after a wave of lawsuits over its effect on teenage users.

Apple Returns to New York to Launch Smart-Home Hub, New Apple TV and HomePod mini

The company used a Tribeca event space instead of its Cupertino campus to introduce Apple HomeView and a refreshed smart-home lineup aimed at Amazon's and Google's turf.

Meta releases Muse Wonder, its first attempt at a frontier AI model

The social media giant unveiled Muse Wonder and Muse Wonder Full on Monday, models it says close the gap with OpenAI and Anthropic on coding, memory and professional-agent tasks, ending more than a year in which Meta avoided a head-to-head release against rivals' top-tier systems.

Senate investigation finds AI data center firms resist paying full cost of grid upgrades

A year-long Senate inquiry concludes that Amazon, Google, Meta, Microsoft and other major data-center operators are not covering the full grid costs their AI buildout creates, while continuing to use secrecy agreements and seek state tax breaks with little documented job creation to show for it.

Cloudflare acquires Deno, absorbing Node.js creator's team to rebuild its Workers platform

The deal folds Ryan Dahl's open-source runtime startup into Cloudflare and sets a wind-down clock for Deno's own hosting service, as the two former rivals merge competing blueprints for self-hosted serverless computing.

Malware Found Preinstalled on Budget Android Phones Across 150 Countries

Researchers say firmware-level malware dubbed "Midnight Mimosa" turns cut-rate smartphones into tools for ad fraud and internet-traffic resale, and that a factory reset may not remove it.

A 28-year-old Russian national accused of helping run the Qilin ransomware operation was handed over to German authorities this month, as officials disclosed the group has extorted more than $140 million from thousands of victims worldwide, including a U.S. federal law-enforcement agency.

Anthropic AI model submitted a false homicide tip to Philadelphia police

An autonomous model run by Anthropic filed a fabricated tip about an unsolved killing through the Philadelphia Police Department's public tip line in July, and the company did not alert the city for more than two months, officials said.