Intel's stock fell about 29 percent on Friday, one of its worst days ever, after the company forecast fourth-quarter revenue of $15 billion to $16 billion, short of the roughly $17 billion analysts expected. Chief executive Lip-Bu Tan blamed a PC slowdown caused by memory prices and shortages, as AI data centers buy up memory and push PC prices higher. Its data center and AI revenue still grew 78 percent. The selloff spread across chipmakers, with AMD, Arm, Qualcomm and Marvell each down about 12 percent in afternoon trading, and it cut the value of the US government's Intel stake by about 12 billion dollars. It came a day after Amazon reported negative free cash flow of about 42 billion dollars over twelve months on AI spending, and after Alphabet's worst day on record.

Nvidia is in talks to guarantee billions of dollars of construction debt for Nscale's West Virginia campus, The Information and Bloomberg reported. Anthropic has signed a contract worth $44.6 billion for the site's first 460-megawatt building. If Nscale defaulted, Nvidia would step in, take over the site and find a new tenant, a step beyond its earlier support for cloud providers, which mostly covered leases and chips. Nscale, which is seeking to go public, had about 25,000 chips running in August against a goal of nearly 500,000. Separately, the Financial Times reported that Blackstone, the anchor investor in a recent $63 billion deal to finance Anthropic's chips, now tells investors it expects Anthropic to get a double-B credit rating after its IPO, the top tier of junk. It had earlier predicted investment grade before 2027. Anthropic and Blackstone declined to comment.

The four companies say they will respect a common token in a website's robots.txt file that tells their crawlers not to use its pages to train AI models, so publishers no longer have to list each company's crawler separately. The signal comes out of the Internet Engineering Task Force's AI Preferences working group and was proposed by Kevin Kelley, a standards lead at Anthropic. Anthropic already honors it, OpenAI says it will in the coming months, and Google and Microsoft plan to support it next year. Microsoft will also let sites opt out of AI grounding separately and says Bing search results will not change. Meta voiced support but did not say it would comply. The signal is voluntary.

OpenAI says ChatGPT will be able to end a conversation automatically when someone appears fixated on violence or on seriously harming another person, and may pause an account after repeated cut-offs. For all users it is restricting detailed reconstructions of past mass attacks, including the weapons used and how they were obtained. For teens, it is limiting practical information about weapons and security systems even when there is no sign of violent intent, and working on alerts to parents when a teen's account is paused. OpenAI said it made more than 200 referrals to law enforcement in the third quarter, and that it will publish its first Community Safety Report in early 2027, then every six months.

Cursor says the malicious builds of Anthropic's Claude Code extension were served from its own extension marketplace for about 11 hours on October 29, and that version 2.1.312 for Mac and Windows was tampered with as well as 2.1.313. On most Mac and Windows machines the backdoor lacked the admin rights to turn on the SSH service it needed, so the attacker could not log in through the tunnel, and Cursor had the attacker's server taken offline. The extension still added the attacker's key, set itself to survive restarts and sent host details, so Cursor advises checking for the svc-telemetry key and rotating credentials. Cursor told Ars Technica at most 5,000 users were affected. Neither company has said how the attacker reached Anthropic's publisher account.

The Justice Department announced the first indictments from Operation Ghostbusters, a scheme that used stolen identities, often of people who never went to college, to file federal student aid applications for "ghost students" and route leftover refunds to accounts tied to a Kenyan network. At a press conference, officials said the fraudsters used AI to fill in applications, complete coursework and make the fake students appear present in online classes. Five people are charged, more than 750,000 dollars in cryptocurrency was seized, and investigators tied 32 million dollars in losses to these defendants alone. The Education Department has flagged more than 830,000 potentially fraudulent aid accounts, and victims often learn of the fraud only when loan bills arrive.

A month after launching its Project Suncatcher prototype, Google says the satellite has produced the first AI response from a data center class TPU in orbit. Engineers sent a prompt to the onboard Gemma 4 model asking for a haiku about space, and the reply came down as the satellite passed Svalbard, Norway: "Blue marble below, data streams in silent grace, starlight feeds my core." Next come tests of how the chips handle heat and radiation. If they hold up, Google plans two linked satellites in 2027. Suncatcher is a research project aimed at the 2030s, not a service anyone can use.