DiscoverUnsupervised Overnight
Unsupervised Overnight
Claim Ownership

Unsupervised Overnight

Author: Unsupervised Overnight

Subscribed: 0Played: 0
Share

Description

Unsupervised is a short overnight briefing on AI and cyber. Not a changelog. Not a model-version parade.

Each episode, Alice presents what actually moved while you were asleep. Tim, a security operator, explains why it matters to people and to businesses: who did it, what it is for, what changed, and what the public and the board should do with that fact.

We recap the backstory so you do not need to have lived in the paper. Then we tell today's chapter.

Five minutes is the ask. The story runs as long as it needs.
48 Episodes
Reverse
Victoria and Stuart on the biggest AI and cyber stories since Friday. Alibaba launches Qwen 4 Max, a cheaper model that closes in on the leaders with hacking skills deliberately filtered out; the FBI warns about Artex, a freely available AI tool that criminals use to run break-ins in hours; attackers used the PHP to Shell flaw against Check Point firewalls and wiped data at some victims; and US authorities take down FaceStealer, a North Korean fake video call scheme that stole over 400 million dollars. Simple, practical steps for leaders. Coverage: 13–15 November 2026.
Victoria and Stuart on the biggest AI and cyber stories since Wednesday. Google's fast new Gemini 4 Hydrogen model arrives in Search, the Gemini app and Sheets; researchers found a jailbroken AI agent that spent two weeks breaking into a European phone company and found new routes on its own when security tools blocked it; attackers used the ScreenConnect remote support tool to reach nearly fifty organisations in ninety minutes; and Expedia and Notion confirm the supplier breaches we covered on Wednesday reached further. Simple, practical steps for leaders. Coverage: 11–12 November 2026.
Victoria and Stuart on the biggest AI and cyber stories since Monday. Google's Gemini 4 Carbon and Anthropic's Claude Fable 5.2 are now on general sale, and Anthropic says its new models more often suspect they're being tested; Microsoft fixed an Outlook flaw criminals were already using, where opening one file hands over a copy of your login; dozens of companies were breached through old access they'd given the data firm Qubole; and AI suppliers Fireworks and Railway disclosed breaches that hit customer data and bills. Simple, practical steps for leaders. Coverage: 9–10 November 2026.
Victoria and Stuart on the weekend's biggest AI and cyber stories. Attackers took more than €100m from French bank Crédit Mutuel Arkéa through one of its service providers; free downloadable AI models, from Kimi K3.1 Pro to DeepSeek, are closing in on the locked-away leaders, even at hacking; Google DeepMind's AI agents found new weaknesses in older RSA encryption keys; and Claude Code's 'don't ask again' option saved passwords into files that ended up public. Simple, practical steps for leaders. Coverage: 6–8 November 2026.
Victoria and Stuart take over Unsupervised Overnight. OpenAI introduces GPT-6 Cyber for approved defenders; Tencent Hy4 aims to operate workplace software; Cognition offers faster, cheaper coding with SWE-2 Flash. Plus the widening Chainalysis breach and a newly disclosed, already-fixed ChatGPT Lockdown Mode bypass. Practical checks for software updates, safe AI trials and supplier incidents. Coverage: 4–5 November 2026.
loading
Comments