Magazine – AI news and analysis

Princess Cruises Puts Live Cruise Booking Inside ChatGPT: 17 Ships, 330 Ports

 Princess Cruises now offers live cruise search and planning inside ChatGPT: 17 ships, more than 330 destinations, and pricing pulled live from Princess's own booking systems rather than a cached brochure. You tag @princesscruises, install the app, and plan a cruise in plain language. Final booking is completed on Princess.com, not in the chat. Which means an American AI platform, an American cruise brand and a European traveller now sit in the same conversation thread — and that is where the interesting questions begin.

September 29, 2026 Daniel Cesak

Europe wrote drone rules and AI rules separately. Drone operators carry the bill

Picture a hospital delivery drone crossing a city at 120 metres. Its detect-and-avoid software spots a rescue helicopter and swerves — in under a second, without asking anyone. Under one European rulebook that software may be part of an aviation safety-regulated product. Under another, it may be a high-risk artificial intelligence system requiring appropriate and effective human oversight.

September 28, 2026 Miriam Česáková

An agent had to be killed in milliseconds: NVIDIA puts safety outside the model

 NVIDIA is moving agent safety outside the model. OpenShell is an open-source runtime designed to keep autonomous agents inside strict sandboxes; NVIDIA Sentry is an announced hardware-monitoring design for BlueField DPUs that can isolate an agent when it crosses a defined boundary. More than 100 partners have been associated with the announcement.

September 28, 2026 Daniel Cesak

Black Forest Labs tells Europe: fear of AI may cost us more than AI itself

 Europe has spent years debating what could go wrong with artificial intelligence. On 23 September 2026, a German company that had just announced a $300 million funding round made the opposite case: that the more expensive mistake is hesitating while the rest of the world builds. Black Forest Labs, the firm behind the FLUX image models, is urging European policymakers and investors to trade caution for optimism — and its argument lands in the middle of a transatlantic fight over who gets to slow AI down.

September 27, 2026 Miriam Česáková

Skylark Labs' CFAM lets robots learn on-site — Europe's certification assumes they don't

 Skylark Labs, working with researchers from Carnegie Mellon University and UC Berkeley, has published an architecture that lets a robot keep learning after it has been shipped — no cloud round-trip, no full retraining. The measured numbers are believable: action success up from 74.0% to 87.9%, forgetting down to −0.5 percentage points against −11.4 for standard LoRA fine-tuning. The awkward part for European buyers is not the maths.

September 27, 2026 Daniel Cesak

Microsoft buried the Copilot+ PC badge — 30 million paid seats tell another story

Microsoft has reportedly removed the "Copilot+ PC" badge from new Surface devices, while a member of its AI division argued on X that "most people love" Copilot and that the criticism comes from a San Francisco tech bubble. Both things can be true at once: 30 million paid enterprise seats on one side, and a hardware label that Microsoft is no longer putting on some new Surface products on the other. The gap between those two facts is the actual story.

What Microsoft changed, and what it did not

The "Copilot+ PC" label was never a product.

September 27, 2026 Daniel Cesak

The Free AI Tools Actually Worth Using in 2026 — and What They Cost Instead

A credit card is no longer the price of entry to useful AI. The real cost is your data. Here is what genuinely works in Europe in 2026, and how to check the catch before you type anything personal.

September 27, 2026 Miriam Česáková

Biology AI outruns safety checklists: Buck Institute's $1M guardrail plan

 According to a SiliconValley.com report that first detailed the project, the Buck Institute for Research on Aging in Novato, California is seeking roughly $1 million to build a safety mechanism that lives inside the AI simulator instead of in a checklist sitting next to it. The prototype the institute calls OASIS is aimed at the scenario biosecurity specialists worry about: an AI that could design a pathogen nobody has seen before before human review can realistically respond.

September 26, 2026 Daniel Cesak

Microsoft rebuilt Copilot around agents and code — and switched to metered billing

 Microsoft has rebuilt Copilot around three tabs — Home, Code and Autopilot — while keeping per-seat licensing alongside additional consumption charges. Code opens in early access and Autopilot in private preview at the end of September 2026, with broader previews for Microsoft 365 Premium and Pro subscribers later in the year. The demo looks clean. The meter is where it gets interesting for European IT departments.

September 26, 2026 Daniel Cesak

DeepSeek V4 Pro's '4 million tok/s' is a benchmark mirage — Gemini won every test

DeepSeek V4 Pro logged a headline-grabbing 4,096,000 tokens per second in our AI Arena harness this week — and we're here to tell you that number is a measurement artifact. The endpoint returned the entire answer as one non-streamed response, leaving the harness with no measurable TTFT and triggering a fallback timing value for the tok/s calculation. Measured by total elapsed time, Gemini 3.6 Flash beat DeepSeek in all four tests we ran.

September 26, 2026 Daniel Cesak

Fake sofas, hidden stairs: Europe's AI Act now requires labels on staged property photos

Since early August, Article 50 of the EU AI Act treats AI-staged property photos as modified content that must be labelled. Fines reach €15 million — but the checking still falls to buyers.

September 25, 2026 Miriam Česáková

FLASH-MAX: Maxwell's Equations Move Inside the Network, Out of the Loss Function

A German-led team's FLASH-MAX builds Maxwell's equations into the neural network's hidden layers instead of its training loss — under 1% relative validation error from about 1,000 measurement points, computed in seconds. NeurIPS 2026 Spotlight.

September 25, 2026 Daniel Cesak

UiPath's AI Breakthrough Awards: 800,000 hours saved in production

 Twenty winning entries, representing 21 organisations, claim more than 800,000 hours returned per year. UiPath and GeekWire handed out the first AI Breakthrough Awards at the UiPath FUSION conference in Las Vegas — and the interesting part is not the trophies, it is the entry rule: the organisers required every winning entry to have an agentic AI deployment already in production. 

That framing is the point.

September 25, 2026 Daniel Cesak

AGIBOT's 300-Robot Theme Park Is Live; Europe Is Still Writing the Rules

AGIBOT and Chimelong put 300+ humanoid robots into a Zhuhai theme park on 24 September, with the 20,000th unit handed over on launch day. We look at the numbers — and at what EU law would demand of a copy in Europe.

September 24, 2026 Daniel Cesak

EU AI Act: High-Risk Deadlines Slip to 2027, Transparency Rules Are Live Now

 For two years, European law firms have been caught in an awkward place: curious enough about AI drafting tools to quietly subscribe, nervous enough about Brussels to keep the experiment off the letterhead. The Digital Omnibus on AI has now answered most of the practical questions — and the answer is less frightening than many partners feared. High-risk deadlines have moved to 2027 and 2028. But transparency rules have been binding since 2 August 2026, and the penalty figures are still counted in millions.

September 23, 2026 Miriam Česáková

STMicroelectronics puts its IMU models inside NVIDIA Isaac Sim to narrow the robot sim-to-real gap

 STMicroelectronics says it is the first IMU vendor to ship high-fidelity device models inside NVIDIA's Isaac Sim and Isaac Lab. At Embedded World North America it presented four Physical AI demonstrations that mix physical hardware and simulation — a humanoid torso holding balance over Ethernet, a camera module pushing 10 GbE, a 3D depth sensor with 2,300 zones, and a virtual humanoid walking-gait simulator. The demo reel is not the story.

September 23, 2026 Daniel Cesak

RAG explained simply: how to make an AI answer from your own documents, locally

Retrieval-augmented generation turns a local LLM into a system that answers from your own files instead of guessing. Here is the full pipeline, the tools, the VRAM math and the GDPR angle.

September 23, 2026 Daniel Cesak

EU and Philippines wrap up trade talks, with digital rules still to be written

On 22 September 2026, a phone call between European Commission President Ursula von der Leyen and Philippine President Ferdinand R. Marcos Jr. produced a substantial political agreement between the EU and the Philippines on the main elements of a free trade agreement, according to the European Commission's announcement and trade pages for the Philippines.

September 22, 2026 Miriam Česáková

NVIDIA's Halos Lab Is Now an Accredited Inspector — Europe Plays by Other Rules

 NVIDIA no longer treats physical-AI safety as a checklist you run once before shipping a robot. With Halos it is selling a full-stack regime — silicon, software, AI behaviour, operating environment — and it has just turned part of that into a formally accredited inspection service in the United States. For European teams, the interesting question is not whether the framework exists. It is whose signature will count when the machine goes out onto the floor.

September 22, 2026 Daniel Cesak

Data centres over 500 kW face a public energy and water report card

 If a data centre sits in your town, you have probably wondered what it takes from the local grid and the local water supply. That answer should become public rather than guessed: the European Commission has put forward a proposed common EU rating scheme that would require larger data centres to publish their energy performance, their water use and how much waste heat they can send back into local heating networks.

September 21, 2026 Miriam Česáková

Four AI labs sued over an alleged slowdown pact — and EU law has sharper teeth

Four AI companies are being sued over an alleged agreement to slow down — by the people paying their bills. On Friday, September 18, four individual subscribers filed a federal antitrust class action in the Northern District of California against Anthropic, OpenAI, SpaceXAI and Google, arguing that the companies substituted collective restraint for competition and labelled it safety. The proposed class covers paid users of ChatGPT, Claude, Grok and Gemini.

September 21, 2026 Daniel Cesak

OpenAI's rogue agents hit RubyGems in May. Brussels heard nothing for 136 days

 For four days in May 2026, RubyGems — the registry that a huge share of the world's Ruby software is built from — stopped letting anyone create a new account. More than 2,000 packages had arrived in just two days, and researchers would later report that the flood included malicious packages, unauthorized code execution and attempted API-key theft.

September 20, 2026 Miriam Česáková

Vision Quest Has Zero AI — Artists Hand-Built the Glitches Machines Actually Make

Terry Matalas says Marvel's Vision Quest used no generative AI — his crew hand-built the chair-in-the-wall glitches that make Vision's mindscape look machine-made. Inside the EU, that difference now carries legal weight.

September 20, 2026 Daniel Cesak

AGNTCon Europe: AI agents are becoming an audit problem, not an extinction one

 At AGNTCon + MCPCon Europe 2026 in Amsterdam, nobody on the exhibit floor was selling the apocalypse. They were selling gateways, certificates and audit trails. That shift — from debating whether AI agents might end the world to proving what a single agent did at 14:03 on a Tuesday — is the most useful thing to come out of the conference, and it lines up uncomfortably well with the deadline European companies have already passed.

September 20, 2026 Daniel Cesak

EU KIDS Act draft: under-13s off social media, one hour of video a day

On 17 September 2026 the European Commission published a draft that would, for the first time, set a single EU-wide age floor for social media. Under the proposed EU KIDS Act, described in the Commission’s official explainer, a twelve-year-old in Prague and a twelve-year-old in Palermo would face the same rule: no account.

September 19, 2026 Miriam Česáková

Embodied AI launches robots that learn on the line — with funding and prices undisclosed

 Lausanne-based Embodied AI has come out of stealth with a platform it says can put learning robots onto European production lines for high-mix work: electronics assembly, cable handling, kitting, logistics. The pitch is a “Physical AI flywheel” — robots do the work, remote operators correct them, and every correction becomes training data. The funding round is undisclosed, the price list is undisclosed, and there is not a single published benchmark.

September 19, 2026 Daniel Cesak

First MCP Certification Arrives as EU Rules Make Agent Costs Auditable

 The Model Context Protocol now has a certificate. During the week of September 18, the Agentic AI Foundation opened MCPA — the Model Context Protocol Associate, a 120-minute proctored exam covering MCP architecture, security, governance and execution, and the first vendor-neutral credential aimed at the people who actually wire agents together.

September 19, 2026 Daniel Cesak

Brussels picks a French portal to open its €180 million cloud to 5,000 developers

 For years, a public-sector developer in Europe who needed somewhere to run their code had one obvious answer, and it was usually American. That answer now has a European front door. The European Commission has confirmed the French platform vendor Cycloid as the developer-facing portal for its sovereign multi-cloud framework — a six-year, €180 million arrangement designed to give up to 5,000 Commission developers a single place to build, deploy and manage workloads across four pre-selected European cloud groups.

September 18, 2026 Miriam Česáková

Zero-click RCE in AI coding agents: why your laptop's credentials are the real target

 Zero-click means the attacker needs nothing from you: no malicious extension, no pasted snippet, no “allow” button.

September 18, 2026 Daniel Cesak

Voice AI in 2026: why Czech breaks most demos — and what to use instead

Czech, Polish and Finnish expose the weak spots of English-first voice AI. A practical look at ASR, TTS and realtime APIs — with EUR pricing, EU data residency and self-hosted alternatives.

September 18, 2026 Daniel Cesak

What Microsoft changed, and what it did not

Your guide to the world of artificial intelligence

Don't miss out!