Anthropic's draft IPO filing shows $16.3B revenue and an $80.7B loss, WSJ reports

The Wall Street Journal has seen a draft prospectus from Anthropic, the AI lab behind Claude, ahead of its planned stock listing. Revenue in the first half of 2026 was $16.3 billion, up from $1.1 billion a year earlier. About 20% came from two customers. The net loss was $80.7 billion, mostly from accounting charges. Bloomberg, citing the WSJ, says Anthropic spent about $9 billion on computing and infrastructure in the half, up from $2.3 billion, and could spend $750 billion more in the coming years. TechCrunch reports Anthropic is struggling to get enough computing power from cloud providers that increasingly see it as a competitor, ahead of a planned November IPO.

OpenAI buys Orchestra Bio, a startup making AI agents for drug development

OpenAI announced it has acquired Orchestra Bio, a four-year-old startup. Its AI agents help biotech companies with the day-to-day work of drug development, from tracking finances to sorting out clinical trial schedules. OpenAI says the team and technology will join work on Rosalind, its AI model for life sciences research, and the related Rosalind Workbench tool. Terms were not disclosed. TechCrunch says the startup had reportedly raised about $20 million as of 2025.

Chinese AI lab DeepSeek opens up DeepInfer, the software it uses to run its models

DeepSeek has released DeepInfer as open source, Bloomberg reports. DeepInfer is the system that runs DeepSeek's production AI services. It starts with support for DeepSeek's V4.1-Flash model. Bloomberg says the release could make DeepSeek's technology cheaper and easier for businesses to run. The code is on GitHub, and the story drew more than 2,000 points on Hacker News.

Google's Gemini assistant in cars can now use your Gmail, Calendar and Photos

Google is rolling out Personal Intelligence for Gemini on Android Auto now, and to cars with Google built-in in the coming months. With it, Gemini can draw on Gmail, Calendar, Maps, Photos, YouTube and Search. Google's examples include checking a flight's status from a Gmail confirmation, or saying "take me to my doctor's appointment". The feature is opt-in, and personal answers only appear when your phone is connected to the car.

SpaceX's revenue jumps 161% on AI, and Musk says a Tesla merger is being studied

SpaceX reported third-quarter revenue of $14.58 billion, up 161% from a year earlier and above the $12.72 billion analysts expected, CNBC reports. AI revenue rose 830% to $7.56 billion, and the AI unit made $431 million in operating income after a $1.7 billion loss a year ago. Connectivity revenue, which includes Starlink, rose 69% to $5.2 billion, the WSJ reports. TechCrunch puts net income at $4.4 billion. Spending on data centers and new technology reached $19.6 billion, up from $6.9 billion. On the call, Elon Musk said Tesla and SpaceX have formed special committees of independent directors, at shareholders' request, to discuss a possible merger. He called it a good idea but said it was not guaranteed, and that he sits on neither committee.

Meta details Prometheus, its Ohio AI supercomputer headed for about 1 million GPUs

Meta says the first phase of Prometheus, in New Albany, Ohio, is complete. It was used to train Meta's Muse models, including Muse Wonder. Prometheus now has hundreds of thousands of GPUs across several linked buildings. It runs two Nvidia platforms, based on GB200 and GB300, and a third based on Vera Rubin is arriving soon. Meta says its pre-training runs kept more than 85% of their compute doing useful work.

Browsers block eight compromised web security certificates from issuer eMudhra

Eight intermediate TLS certificates from eMudhra were found to be compromised, Ars Technica reports. Their private keys were only 3 to 8 bits long. Intermediate certificates like these can be used to create valid-looking credentials for any server on the internet. Browser makers and major apps have blocked them. Cloudflare says it learned of the problem on October 19 and has patched its services and apps. It says it has found no active exploitation on its services.

Shared chats from Meta's AI assistant Muse expose users' personal and business secrets

Ars Technica reports that when people share conversations from Meta's Muse app, those chats can leak personal and business secrets. Researchers at the security firm Claroc say they found more than 6,800 Muse links through public DNS records, The Verge reports. Meta has paused the sharing feature while it looks into the privacy problem.

Attackers set fridges to defrost in hundreds of French supermarkets via open controls

On the night of October 11 to 12, attackers pushed refrigeration systems in hundreds of French supermarkets into defrost mode, and stores threw away food. Reports point to controllers made by Danfoss, a Danish industrial company. IPinfo's scans found 1,331 Danfoss fridge dashboards open to the internet in France days before the attack, and 85% ran software with known flaws. France's cybersecurity agency says attacks are probably still ongoing.

Google Cloud outage in London knocks out digital bank Revolut for hours

Google says Cloud services in its London region were having problems starting late Monday, Pacific time. Revolut says its major outage was caused by Google Cloud, and its online services were down for several hours into Tuesday, Bloomberg reports. Neowin reports that other UK banks and services were also hit.

Everything else