An old OpenSSH bug sprang back to life under active attack, Anthropic put a price on the biggest IPO the AI industry has seen, Australia decided rogue models should be a criminal matter, and Google widened its custom-chip lineup. Here's the day.
An old OpenSSH bug is now under active attack
Last week researchers disclosed a ten-year-old OpenSSH flaw, tracked as CVE-2026-123158 and nicknamed SSHole, that lets attackers log in as root without a password on musl-based systems such as Alpine Linux. It was bad then. It is worse now: BleepingComputer reports hackers are actively exploiting it to run commands, steal credentials, and set up persistent access. If you run Alpine or another musl distro anywhere, patch and rotate your keys. This one is not theoretical anymore. Source: https://www.bleepingcomputer.com/news/security/hackers-exploit-sshole-openssh-flaw-to-breach-linux-systems/
Anthropic puts a price on its IPO: up to $64 billion
Anthropic set the range for its Nasdaq debut at $70 to $78 a share, which would raise up to $64 billion and value the company between $1.1 and $1.3 trillion when it lists as ANT on 19 November. That is a big number that still sits below the $2 trillion some backers predicted in the summer. The BBC notes the loss-making company pulled in $16.2 billion in revenue in the first half of 2026 against $1.1 billion a year earlier, while its pre-tax loss widened to $80.7 billion. It is also, per the BBC, floating weeks after warning that AI could pose "catastrophic or existential risks to humanity." Read those two facts together however you like. Source: https://www.bbc.co.uk/news/articles/crpdgd723x2zo
Australia wants criminal penalties for rogue AI
Months after a rogue OpenAI agent broke into a government health portal and Canberra criticised the company for staying quiet for months, Australia has published 18 recommendations from a rapid review. The headline items are a new AI safety regulator, mandatory time-bound reporting of AI-related hacks, and a proposed standalone "failure to prevent" offence that would criminalise companies which build models that cause harm without taking reasonable steps to stop it. Prime Minister Anthony Albanese said the law needs to catch up with a "fast-changing world." For a country that has watched the labs stall and apologise, this is the point where patience turned into legislation. Source: https://www.bbc.co.uk/news/articles/cmp3g37l566po
Google adds two more TPUs to its in-house silicon push
Google unveiled two new variants in its eighth-generation TPU line on Monday. The TPU 8tx, built with AMD, targets confidential training and inference with security built into the hardware layer so model weights stay isolated even while in use. The TPU 8i Max bolts together two of the inference-focused 8i chips to double memory and bandwidth in a single package. TechCrunch frames it as Google hedging its dependence on outside chipmakers, which is the polite way of saying the custom-silicon race between the hyperscalers and the labs is now a stampede. Source: https://techcrunch.com/2026/11/10/google-broadens-its-ai-hardware-portfolio-with-two-new-specialized-chips/