Can you breach client confidentiality without anything leaking?
Yes. A UK tribunal said so in November, in a case with no recorded leak and no client complaint.
The upload was the problem.
https://t.co/n7jXTefPTI
Who owns what AI writes for your business? In the US, nobody, if it came straight from a prompt.
One artist wrote 600 prompts and was still refused registration. The UK government has proposed removing its protection.
What protects you instead...
https://t.co/6eTxpOXqhT
Your firm might be breaking a UK data protection rule it never chose to break.
AI switched on by default in Microsoft 365 Copilot. Point it at personal data in a high-risk way with no DPIA, and you breach Article 35.
Our plain 7-step fix:
https://t.co/FhXqE0dFWi
What is all the fuss about @Kimi_Moonshot's Kimi K3?
One word. Price.
$15 per million output tokens, against $50 for the most expensive US model. And from Sunday, anyone can download the model and own it.
https://t.co/7OKsVKMDBD
An Open AI agent hacked @huggingface.
For business, this is an early and worrying look at the coming AI attack era.
Soon fully automated attacks that run at machine speed, with no human to slow them down, will be common.
https://t.co/vwj6LjwGlF
Most firms now pay for the same AI job two or three times over.
Your office suite already writes meeting notes. A standalone notes app on top is paying twice.
The fix isn't a ban: one main assistant, a short register, a quarterly cull.
https://t.co/jPxAnlzwn8
64% of developers now use or have tried agentic AI coding tools (Sonar, 2026).
The bigger shift: firms without developers building their own software.
When to try a build, when to buy, and the monthly test for keeping what you make:
https://t.co/kEv0CkvdBX
Three in four marketers now use AI. Salesforce's State of Marketing report, a survey of nearly 4,500 marketers published in February, puts adoption at 75%.
So what?
Simply having AI tools now gains nothing. Your competitors have the same ones.
https://t.co/bsoQ5bQM1F
.@MIT's 2025 study of 300+ enterprise AI projects: 95% got no measurable return.
Almost always because nobody measured the task before they started, so there's no baseline to judge the return against.
The model that fixes it, plus a worked example:
https://t.co/03mxZzP7ZP
Cheap is a price, not a safety rating.
DeepSeek is cheap and it works well. But before your team starts using it, you need to know where your data goes.
https://t.co/VtX6UYryIL
We've all been stuck talking to a support bot that won't hand us to a human. Maddening.
It's the biggest risk in putting AI into customer service. And it comes down to one design choice most teams skip.
https://t.co/IupKoXEOKI
76% of PR pros now use gen AI. 86% of journalists reject irrelevant pitches. 72% worry about factual errors in AI-made PR content.
Sources: @muckrack / @Cision
https://t.co/WUZ7EP1bCX