Tech RSS Digest — 30–31 October 2026
Most important
- A tampered Claude Code extension reached thousands of developers through Cursor's marketplace. Anthropic says a malicious actor published a “tampered” copy of the extension it did not build, and Ars reports it was backdoored for nearly 12 hours; Wired describes a supply chain attack putting malicious extensions on thousands of devices. If you use Cursor, check you have the genuine extension.
- OpenAI's latest funding round shrank sharply. Sources say it secured only ~$11B of a $30B target and now seeks ~$20B at a $1.25T valuation, down from $30B at $1.4T, as some investors cut commitments amid the AI selloff.
- Wall Street started pricing the AI build-out as a liability. Alphabet, Meta, Microsoft, Amazon and Oracle now carry $3T+ in AI spending commitments but returned just 9.3% of cash to shareholders in the first three quarters of 2026, a 15-year low; S&P downgraded Meta's outlook to negative on high capex and negative free cash flow, Intel closed down 28.58% on Friday, its worst day since the 1970s, Reddit closed down 29.27%, its worst drop since going public, and neoclouds CoreWeave, Nebius and Nscale have effectively been shut out of the bond market — while Nvidia is in talks to guarantee debt for Nscale's Anthropic data centre.
- Binance's investigation widened. The exchange says a follow-up probe found “additional points of breach” beyond a possibly compromised laptop and that this was a “complex attack”; it has reserved $6.34B+ in company funds for the fallout, while Chainalysis says the hackers used automated money-laundering patterns not seen before and almost all of the $6.3B was taken in the first 30 minutes.
- Google's data centre satellite is online. Its Suncatcher prototype, launched 1 October, is online with Gemma 4 generating a haiku about being in space — the first response from a data-centre-class TPU in orbit, and a sign of where the infrastructure race is heading next.
- The big AI labs agreed on a shared scraping opt-out. OpenAI, Google, Microsoft and Anthropic say they will adopt a new standard publishers can use to opt out of being scraped by AI across multiple crawlers — a cooperative convention, not a binding rule, but a notable shift after years of scraping fights.
- Apple's AI reaches China in December — via Alibaba. Apple confirmed its AI features arrive in China with iOS 27.3, powered by Qwen, while incoming chief John Ternus says Apple doesn't “want to ship AI for AI's sake” in a Wired interview.
- The memory squeeze became the story underneath everything. Sam Altman argues supply-constrained DRAM is better used in data centres than sitting in PCs “being used 5% of the time”; CXMT reported Q3 revenue up 613.2% year on year, Kioxia's operating profit rose 22x year on year, and HP, Asus and Acer moved some PC making back to China as memory costs bite.
- The incident list kept growing. AO3 could be offline for weeks after its security incident, Scribd was breached, Linode reported a security incident, GitHub had an incident affecting Actions, Pages and Webhooks, Bunny CDN reported a security incident, Northflank was breached, and Applebee's exposed customers' personal data after leaving a database password in its app.
- Regulators moved on several fronts at once. France is getting a download tax for content providers (€0.055/GB) that may kill the VideoLAN non-profit; New York banned AI chatbots from pretending to be human with children; the UK CMA opened a probe into whether Meta imposed exploitative terms on publishers and advertisers for AI training content; and Trump declared robot imports a national security threat.
Other notable stories
- Salesforce's M&A spree continued with a reported $375M deal for defence tech startup Rune, which makes AI military logistics software; Anthropic is expected to receive a high “junk” rating from credit agencies following its IPO, and Anthropic renamed its Max plans and removed the 5-hour limit for verified Ultra accounts.
- Baseten is raising ~$2B in two tranches valuing it at $26B and $30B, up from $13B in June; Cognition weighs funding at up to an $80bn valuation; Databricks and SAP expanded their deal to put AI on SAP data; Cognizant put 25,000 staff behind OpenAI; and Thrive Holdings agreed to buy ADT for $7.6bn in cash.
- Amazon can now borrow up to $80B against the AI hardware it buys; Google is in talks with the DOD on securing data feeding Golden Dome missile-defence software; Meta settled a social media addiction lawsuit, averting trial; and DoorDash said it has only received “several hundred” orders from AI agents so far.
- OpenAI tightened limits on weapons-related information ChatGPT provides teens; House Democrats criticised Justice Dept. support of OpenAI in a copyright suit; Meta, TikTok and X lost their UK court fight over Ofcom data demands; South Korea fined Google and Apple $47.6m over app store payments; and Volkswagen lost €5.8bn in a quarter as Porsche, China and tariffs bit.
- AWS items in the window: Bedrock API support and Cross-Region Inferencing for OpenAI models in GovCloud, Web Search on Kiro in GovCloud, MSK now supports TLS 1.3, CloudFront viewer ISP and post-quantum TLS details, Connect email draft saving and flow versioning/aliasing, WAF automatic SDK integration for CloudFront, and RDS for Oracle dropping automatic minor upgrades for some options.
- Also in the window: Google disclosed PageBreak, an AI security agent that found 500+ XSS bugs; Netflix treats GPU reliability as a distributed systems problem; Kubernetes Confidential Containers gained Fortanix attestation and key management; LocalStack for Azure entered public preview; GPT-6.1 Terra vs. GPT-6 Astra compared; Ford says its new EV drivetrain is 30% more efficient than BYD's; and this week's Self-Host Weekly landed.
Hacker News picks
The front page was dominated by the supply-chain scare and the incident wave: backdoors in official Claude Code extensions, Tell HN: Scribd breached, Linode security incident, GitHub incident with Actions, Pages, and Webhooks and Bunny CDN security incident, alongside lighter fare — a working Nintendo in ~1GB of Qwen weights, Sanstun, a sans-IO WireGuard implementation in Rust, Infinite Mac: BeOS, tools Hacker News keeps asking for, Digital Minimalism, AI isn't replacing SREs, it's deskilling them, switch lowering in GCC and how long do you need to hold down the power button to turn off a PC?
Overall theme: The AI boom's financing strain showed up everywhere at once — a shrunken OpenAI round, a negative outlook for Meta, a record-bad day for Intel, and neoclouds shut out of the bond market — while the week's other lesson was supply-chain trust, with a tampered developer extension and a string of platform incidents landing in the same window.
Sources: 256 items from AWS, Cyber, Software Updates, Tech Dev, Tech Leadership and Technology, plus Hacker News, published 30 October 07:05 – 31 October 07:10 Toronto time.