London mayor halts Metropolitan Police's £50m Palantir AI deal

London mayor Sadiq Khan has blocked the Metropolitan Police from signing a £50 million contract with Palantir, the controversial US surveillance firm, for AI tools to support criminal investigations.

The Met had argued Palantir was the only company able to supply what it needed, but the mayor stepped in following widespread concern from civil liberties groups about handing highly sensitive police data to a firm with known links to US immigration enforcement and the Israeli military.

For UK founders and public sector operators, this is the clearest signal yet that AI procurement decisions are now a political risk, not just a technical or commercial one, especially where biometric or criminal data is involved. Read more

Scotland's "green data centre" policy is masking real AI emissions

An analysis by charity Action to Protect Rural Scotland, covered by The Guardian, shows Scotland's national policy for encouraging "green data centres" was written in 2022, before the release of ChatGPT, and has no working definition of what "green" actually means.

As a result, over a dozen data centres in Scotland, including an AI Growth Zone near Glasgow backed by £8.2 billion in private funding, could label themselves green while the combined electricity demand of these projects is projected at around 6.2 gigawatts, roughly one and a half times Scotland's entire peak winter usage.

This matters for UK businesses making sustainability claims tied to Scottish cloud or compute infrastructure, and for investors in UK AI infrastructure who have assumed "green" carries real meaning.

Apple adds Anthropic's Claude and OpenAI's Codex directly into Xcode

Apple is integrating both Anthropic's Claude Agent and OpenAI's Codex as first-class coding tools inside Xcode, its developer environment used by hundreds of thousands of app builders worldwide.

Developers will be able to call on either AI to write and edit code, update project settings, and handle tasks inside the same tool they already use to ship iOS, macOS, and other Apple platform apps.

For UK development teams and agencies building on Apple platforms, this reduces the need for third-party AI coding tools and brings AI-assisted development into the standard workflow at no extra cost.

KPMG is touring Silicon Valley every few weeks to find AI startups before they can disrupt its business

KPMG's US chief executive Tim Walsh and his management team are now visiting Silicon Valley regularly to meet early-stage AI startups backed by funds including Andreessen Horowitz and Bessemer, according to the Financial Times.

The goal is to identify companies that could eventually threaten KPMG's core audit, tax, and consulting work, and either partner with them or take strategic minority stakes before they grow large enough to compete.

For UK professional services firms, this is a clear signal: the Big Four are no longer treating AI as a client problem. They are treating it as a survival question, and they are moving fast to own the answer.

OpenAI and Anthropic are now going after the £740bn management consulting market directly

OpenAI and Anthropic are both building advisory and professional services practices to compete directly with the Big Four and top management consultancies, targeting a global market worth around £740 billion, according to The Telegraph.

Both companies are positioning their AI models not just as tools for consultants to use, but as the product itself, capable of running analysis, drafting strategy documents, and advising on decisions with less human time involved.

For UK founders and operators who buy professional services, this is worth watching. The next two to three years will likely redefine what a consultant does, what firms charge for, and which parts of the service stack can be automated.

A voiceover artist says ScotRail is using an AI clone of her voice without her consent

Gayanne Potter, one of Britain's most recognised voiceover artists, says her voice was used by Swedish tech company ReadSpeaker to create ScotRail's AI train announcer "Iona" without her meaningful consent, and she wants it removed immediately.

ReadSpeaker says it has addressed her concerns, but Potter says she only agreed to limited use of her voice data and was never told it would be used to build a synthetic announcer deployed across Scotland's train network.

The case raises a direct question that every UK business using AI voice tools needs to answer: do your contracts with voice talent actually cover the AI use you have in mind, and do the people whose voices you use know what they agreed to?

Summarise anything you do not have time to read

Paste any long email, article, contract, or document into ChatGPT or Gemini and ask: "Summarise this in five bullet points and flag anything I need to act on."

This works on planning documents, supplier proposals, NHS letters, school communications, or any dense text you keep skipping. You stay informed without reading every word.

Prepare for meetings and calls in two minutes

Before any important call, ask an AI: "I am about to speak with [role or type of person] about [topic]. What are the three things I should know and the two questions I should ask?"

You arrive better prepared without researching for 30 minutes.

Turn your rough notes into a clean message

After a call or meeting, paste your rough notes into an AI and ask: "Turn this into a clear follow-up email in plain English."

Edit before sending, but you cut writing time from 20 minutes to 3 minutes.