News and Insights

Practical updates on AI governance, workforce strategy, and digital resilience for EU organisations
Supermarket fridges were scouted first TikTok’s feed design is now the risk The data centre backlash and your AI bill EU bets €525m on chip design Your AI suppliers are heading to Congress The midterms change who asks the questions on US AI regulation Gemini case: chatbots and press rights Italy: email open tracking needs consent Mistral opens Large 4 for cyber defence

Stay ahead of AI change. Get practical updates from Future Prep direct to your inbox.

By subscribing you agree to receive Future Prep news. Unsubscribe any time.

Latest Insights

From the Future Prep blog

AI generated image - a large data centre under construction at the edge of farmland behind a temporary fence, illustrating data centre capacity

The data centre backlash and your AI bill

Opponents won almost every one of 41 local data centre ballot measures in the US midterms. Analysts now expect slower buildout and higher grid costs. For a European firm running AI on US infrastructure, data centre capacity has become a question of dependence. Four questions to put to suppliers.
AI generated image - a witness seated at a congressional hearing table facing lawmakers on a raised dais, illustrating AI vendor risk

Your AI suppliers are heading to Congress

The US midterms put AI suppliers in front of Congress: hearings, records requests and pressure to explain safety decisions and job effects. For a European firm running on their models, AI vendor risk shifts long before any law changes. Four questions to answer now.
AI generated image - the United States Capitol dome in morning light behind bare autumn trees, illustrating US AI regulation after the midterms

The midterms change who asks the questions on US AI regulation

Democrats have won both chambers of Congress, and AI was on the ballot. Hearings and subpoenas come first; laws still need the president’s signature. What the result means for US AI regulation, the states and a European firm with American clients or suppliers.
AI generated image - A row of identical mailboxes opened by a single key, illustrating a third-party breach at a shared supplier

The Third-Party Breach That Ran For Three Years

Cathay Pacific says an intruder used an Amadeus system account to reach booking data from August 2023 to September 2026. The GDPR’s notification clocks start at awareness, not at the breach. What a third-party breach of this kind means for your suppliers and your own clients.
AI generated image - facilities manager hands a contractor a single key while the key cabinet stays locked, illustrating AI agents permissions

AI Agents Need Keys, Not Blank Cheques

Spain’s data protection authority published ten recommendations for people who let AI agents act for them. Read with an organisation in mind, they become four controls and a vendor checklist, plus a rule the AI lead must own: which agents staff may use.
AI generated image - man in a suit looks at a fully enclosed engine under an open car bonnet, illustrating third-party AI assurance

Who Actually Understands Your Third-Party AI?

A Bank of England and FCA survey shows UK financial firms source half their important AI from third parties, yet only 24% report deep knowledge of those systems. Why a contract is not understanding, what DORA expects and four questions to ask. Categories: Governance (primary), Compliance, Strategy

Latest News

Short updates

Supermarket fridges were scouted first

Security firm Lupovis says it logged over 276 probing attempts from 91 actors against Danfoss refrigeration controllers between 20 and 30 October, before supermarket cooling systems in nine European countries were hit in early November (Euronews, 7 November). ENISA had already urged organisations to cut the exposure of such systems. Attribution is open. For an EU organisation, the question is which cooling and building systems face the internet, and who watches them.

TikTok’s feed design is now the risk

On 6 November the European Commission preliminarily found that TikTok’s For You feed breaches the Digital Services Act. In its view TikTok did not assess how the feed pulls users into harmful rabbit holes, and the “not interested” button does too little. It wants explicit preferences weighted more and viewing time no longer used as a key metric. For anyone running a recommender, the question is which metric it rewards and who signed that off.

EU bets €525m on chip design

The Commission has launched the European Chip Design Platform, aiming to mobilise €525 million for chip design start-ups. It offers design tools, IP blocks, manufacturing access and grants of up to €16 million. European firms earn less than 1% of global chip design revenues, according to the Commission. For an EU organisation buying AI, nothing changes this year. The real question is whether Europe can design what it wants to depend on.

Gemini case: chatbots and press rights

In a Hungarian publisher’s case over Google’s Gemini, Advocate General Szpunar told the EU Court of Justice on 3 November that a chatbot’s answer can infringe press publishers’ rights. Two tests decide: did the chatbot reach the article lawfully, without dodging a paywall, and does it reuse the wording so closely that readers no longer need the original? Facts stay free. The opinion is not binding. Does your assistant report news or repeat it?

Italy: email open tracking needs consent

Italy’s data protection authority, the Garante, gave organisations six months to comply with its guidelines on email tracking pixels. That period ended in late October. Measuring who opened which message, to judge or adjust a campaign, now needs prior consent. A pixel identical for every recipient, counting only the overall open rate on anonymised data, needs none. Every pixel belongs in the privacy notice, whatever the email. Which of your engagement scores rest on consent?

Mistral opens Large 4 for cyber defence

Mistral released the weights of Large 4 on 29 October: a model of about one trillion parameters that organisations can download and run on their own servers. Mistral aims it at defensive cyber work, from finding vulnerabilities to patching them. It says built-in guardrails separate defensive from offensive requests, and keeps a specialised cyber variant for vetted partners. Moonshot removed cyber data before its own release. Which approach would your security team rather trust?

Scroll to Top