Daily AI brief · November 5, 2026
One Unwatched PC Let Ransomware Lock 37 Machines
Ransomware reached about 50 machines at a UK school group from one computer nobody was watching. Also: Claude can now connect a whole team's apps at once, and ChatGPT makes animated GIFs.
One unwatched computer let ransomware lock about 37 machines at a UK school group (Bitdefender, BleepingComputer, Infosecurity Magazine) Security firm Bitdefender published its investigation on Tuesday into a 3AM ransomware attack on a group of UK schools. Every step the attacker took came from one computer that had no security software on it. From there they pulled saved passwords out of the backup servers, created two hidden administrator accounts, linked three separate school networks together, and encrypted about 37 of the roughly 50 machines they reached over about ten days in September. Bitdefender suspects the attacker had an AI writing commands in a loop, but says it can't prove it, and it couldn't tell how the attacker first got in. If you run the office, the AI part matters less than the forgotten machine: the old PC in the back room, the reception computer nobody updates, the server a former vendor set up. Find yours this week.
Claude Team admins can now connect an app once for everyone (Anthropic) Anthropic added a way for the admin of a Claude Team or Enterprise plan to switch on a connector, the link that lets Claude read and work in another app like your CRM or file storage, for the whole team at once. Staff get it the first time they sign in, with no setup of their own. Without a single sign-in system, it's a beta where Claude signs people in by their work email, and Anthropic's help page says that check is membership only, not two-step sign-in. So offboarding now has two steps: remove the person from your Claude organization, and remove them from the connected app too. Whoever handles AI for you can see what's switched on today in Claude under Organization settings > Connectors, then pick a connector to see who gets it.
ChatGPT can now make animated GIFs from a prompt or a photo (The Decoder, citing The Verge) OpenAI's new image model, Images 2.6, lets ChatGPT turn a text prompt or an uploaded photo into an animated GIF, then change the speed or fix single frames. It's rolling out to all ChatGPT users on the web, desktop and phone apps, and OpenAI says people now make more than four billion images a week across ChatGPT and its developer tools. The practical part: a small marketing team can turn a product photo into a short moving post for social media in a few minutes, with no designer. Use your own photos, and skip customers' faces unless they've said yes, because a GIF travels further than the photo ever did. Before you post one, play it through and check for warped text or a mangled logo.
Teach it your voice with three old emails (Next step · 10 minutes)
The drafts that come back from Claude often sound like nobody in particular. Fix that once: pick three emails you've sent that sound like you on a good day, take out anything private, and paste them in with the prompt below. Then ask for the reply you've been putting off. Customers notice when a note sounds like the person they know.
> Here are three emails I wrote: [paste]. Describe my writing style in five short bullet points: length, tone, greeting and sign-off, words I use, words I avoid. From now on in this chat, write in that style.
Works the same in ChatGPT and Copilot. The next tip saves it so you don't paste them every time.
What good looks like: five bullets that make you nod, and a draft you'd send with one or two changes. Avoid: using your stiffest emails as samples. Pick the ones customers liked.
Walk the office with a notepad. Fifteen minutes, one list: every computer, laptop, server, printer and anything else plugged into the network, who uses it, and whether it gets updates and runs security software. If an outside vendor ever installed something, a server, a camera system, a time clock, ask them for their list too. You'll probably find one machine nobody can answer for. Update it and protect it this week, or take it off the network. Then check the backup system's sign-in, since that's where the attackers in today's lead found saved passwords. A machine on the list costs nothing; a machine off the list is how 37 got locked.
Try a meeting in two languages: if you're on Google Workspace, Google says its Gemini speech translation in Meet, now covering 63 languages, began reaching business accounts on Google's standard release schedule on November 2 and can take up to 15 days to appear. It's off by default; each person turns it on during the call and hears the other side in their own language. If you have a supplier, a customer or a crew lead who'd rather speak another language, set up a short call and try it, and keep a person checking anything with a price or a date in it. It doesn't work on dial-in calls or in recordings.
Where it's going: AI is speeding up both sides at once, the attacker's commands and your team's setup, so the offices that know every machine and every connection they have will get the speed without the surprise.
- 3AM ransomware at UK schools: [Bitdefender, Nov 3](https://www.bitdefender.com/en-gb/blog/businessinsights/3am-ransomware-hybrid-ai-attack) · [BleepingComputer, Nov 3](https://www.bleepingcomputer.com/news/security/uk-school-estate-hit-by-3am-ransomware-in-suspected-ai-assisted-attack/) · [Infosecurity Magazine, Nov 4](https://www.infosecurity-magazine.com/news/3am-ransomware-hybrid-ai-uk-school/)
- Claude connectors for the whole team: [Claude Help Center: Auto-connect your organization's apps to Claude, Nov 3](https://support.claude.com/en/articles/15537633-auto-connect-your-organization-s-apps-to-claude)
- ChatGPT animated GIFs: [The Decoder, Nov 4](https://the-decoder.com/chatgpt-can-now-turn-text-prompts-and-photos-into-animated-gifs-with-openais-images-2-6-model)
- Google Meet speech translation: [Google Workspace Updates, Oct 22](https://workspaceupdates.googleblog.com/2026/10/real-time-speech-translation-in-google-Meet-expands-to-63-languages-with-automatic-language-detection.html)