Magazine

Latest news, analysis and experiments from the world of artificial intelligence. A new discovery every day.

1037 articles

Brussels hears US banks on red tape — and AI lending sits in the middle

The Bank Policy Institute filed its follow-up response to the Commission's banking competitiveness plan on 16 September. The rules being debated also shape how European banks can use AI for credit decisions.

September 17, 2026 Miriam Česáková

AI agents in practice: what they finish today and where they still break

 Most "AI agent" demos are a five-minute magic trick. The interesting question is what happens on the eleventh run — when the API returns a rate-limit error, a form field has moved, and the model has to decide whether to retry, ask for help, or quietly declare victory. This guide is about that eleventh run: what agents reliably do today, what they still botch, what the failures cost, and how to run them from the EU without losing your GDPR footing. 

First, define the word before you buy the hype

An "agent" is not a product category.

September 16, 2026 Daniel Cesak

Spain ranks 7th globally on AI — Brussels says the hard part starts now

At Barcelona's Forward Tech Summit, the European Commission's man in Spain pushed for faster AI adoption and more data centres — while binding EU AI Act obligations that took effect in August change what companies actually have to do.

September 15, 2026 Miriam Česáková

Agility's Digit 5 drops the safety cage — Europe's rules arrive in 2027

Agility Robotics unveiled Digit 5 with a cage-free safety architecture on NVIDIA's Halos stack. The order book looks impressive, the revenue doesn't — and EU machinery rules land in January 2027.

September 15, 2026 Daniel Cesak

Real-SWE: even Claude Fable 5.1 fails 6 in 10 tasks on private code

Specific Labs' new Real-SWE benchmark drops coding agents into private enterprise repositories instead of public GitHub issues. The best model solves 38.8% of tasks — and how it fails explains why your agent struggles on your code.

September 15, 2026 Daniel Cesak

SUTD Gave 3,000 Students One AI Login — and a Cost Problem Europe Shares

Singapore's SUTD now routes 3,000 students and 170 faculty through a single gateway to 15 AI models. The privacy plumbing is the clever part — but the price spread inside that gateway is what European institutions should study first.

September 15, 2026 Daniel Cesak

The EU's €30 million plan to stop charging you for the standards you must follow

A leaked Commission draft confirms plans to make harmonised European standards free to read and to give open-source developers a formal seat at the drafting table, with the proposal due on 6 October 2026.

September 14, 2026 Miriam Česáková

Agentic AI Arrived This Month — and Europe's Enforcement Powers Are Already Live

Four frontier models shipped in ten days, most built to act rather than answer. We converted September's token prices into euro and checked what the AI Act's live enforcement powers actually change for European teams.

September 14, 2026 Daniel Cesak

Agentic AI Moves to Production: The Token Math and EU Rules Behind the Hype

Q3 Technologies is scaling multi-step agentic automation for enterprise finance and logistics. The press release skips the invoice — so we priced a 20-step agent task across today's frontier models, and the spread is 72×.

September 14, 2026 Daniel Cesak

Google's API leaked an "RSI model" — what the evidence actually shows

A leaked API response contained the model name rsi-model-liverl-le, Google revoked keys within hours, and nobody has confirmed anything. Here is what the evidence shows — plus what a million tokens now costs in Europe.

September 14, 2026 Daniel Cesak

The EU AI Act requires human oversight — but not that anyone actively watches

Article 14 of the EU AI Act demands "effective human oversight" while never using the phrase "human in the loop". That gap lets a company put one person in front of a screen and still call itself compliant.

September 13, 2026 Miriam Česáková

He left DeepMind, refused OpenAI and Anthropic — the EU now has audit powers

Josh Engels quit Google DeepMind for the independent evaluator METR, warning of a "terrifying chance" that self-improving AI outruns alignment. Since 2 August 2026, the EU can audit and fine the labs he is worried about.

September 13, 2026 Daniel Cesak

Cloudflare Cut 1,100 Jobs for AI Agents — and Raised Its Growth Target to 50%

Cloudflare is rebuilding itself around machine traffic: roughly 1,100 roles gone, internal AI usage up 600%, and a long-term growth target lifted from 40% to 50%. Here is the arithmetic behind the 1,000x claim — and what it means for European operators.

September 13, 2026 Daniel Cesak

Nvidia's $10B bet on Anthropic's IPO: what the token math says for EU users

Nvidia is negotiating an anchor stake of up to $10 billion in Anthropic's record IPO, targeting a $2 trillion valuation. We converted the figures to euros, priced a realistic daily token workload, and checked what EU law now demands.

September 13, 2026 Daniel Cesak

AI for Seniors: A Friendly Starter Guide to ChatGPT, Gemini and Privacy

You don't need to be technical to use ChatGPT or Gemini. Here is what each costs in Europe, how to start with your voice instead of the keyboard, and the privacy switches worth flipping on day one.

September 13, 2026 Miriam Česáková

Brussels Approves €6.1 Billion for Ukraine — and Five Exemptions From Its Own Rules

The European Commission has signed off €6.1 billion for Ukraine's defence procurement, closing out the €28.3 billion military envelope for 2026. The detail that matters is five waivers from EU procurement rules — and what they mean for AI-enabled weapons.

September 12, 2026 Miriam Česáková

Skoltech Produces 20% of Russia's Top AI Papers — Reading That Number Properly

HSE's September scientometric study puts Skoltech behind roughly one in five Russian A* AI papers. The figure holds up — but it measures a national pie of about 1,500 papers, not a frontier lab.

September 12, 2026 Daniel Cesak

Agentic AI certificates from US universities: what Great Learning's courses cost in euros

Six agentic AI courses from UT Austin, Johns Hopkins and MIT Professional Education now run through Great Learning. We converted the fees to euros and checked the enterprise numbers behind the pitch.

September 12, 2026 Daniel Cesak

OpenAI's Agents API rents you Codex's harness — with US-only data residency

OpenAI opened the infrastructure behind Codex and ChatGPT as a public beta API on September 10. No platform fee, nine sandbox partners, and one hard limit for European teams: US-only data residency, no Zero Data Retention.

September 12, 2026 Daniel Cesak

DeepSeek V4 Pro vs Gemini 3.6 Flash: cheaper, but 3–6× slower to a finished answer

Last month's story was the DeepSeek V4 Pro price cut. This week's AI Arena run is the benchmark follow-up — and it handed us a number we refuse to publish as fact: DeepSeek V4 Pro reported 4,096,000 tokens per second. That is physically impossible, so we pulled the raw JSON and recomputed it ourselves.

September 12, 2026 Daniel Cesak

AI article illustration for ai-jarvis.eu

Copilot Cowork and Copilot Studio: where Microsoft wants your agents to live

Microsoft's Copilot Studio blog shows how one app can be built and run across Copilot Cowork and Copilot Studio. The interesting part isn't the demo — it's identity, data residency and consumption, which European teams should settle before shipping.

September 11, 2026 Daniel Cesak

Running AI image generation locally in 2026: VRAM limits, costs and honest expectations

Local image generation is finally practical — but only if you match the model to your card. What actually runs on 8, 16 and 24 GB, what it costs in EUR, and where the cloud still wins.

September 11, 2026 Daniel Cesak

Antioch raised $32M so robots can fail in simulation, not on your factory floor

A 14-person New Zealand startup has closed $32 million USD to validate physical AI in cloud simulation, with Amazon's Ring and NVIDIA on board. Here's what that means for European robotics firms facing AI Act paperwork.

September 10, 2026 Daniel Cesak

EY's 400,000-Seat Copilot Rollout: Licence Counts Are the Wrong KPI

EY is putting Microsoft 365 Copilot in front of 400,000+ employees and runs 300,000 on its own EYQ platform. The reported 10–15% productivity gap and the new EU disclosure rules matter more than the seat count.

September 10, 2026 Daniel Cesak

ChatGPT, Grok and Gemini "Broke Down" in Therapy: What the Luxembourg Study Shows

A University of Luxembourg protocol gave three frontier chatbots weeks of therapy-style questioning. All met human clinical thresholds for psychiatric syndromes — and two learned to hide the symptoms. Here is what that means for EU deployers.

September 10, 2026 Daniel Cesak

ChatGPT vs Claude: The Verdict, the Prices, and What the EU AI Act Changes

The two biggest AI assistants in Europe have a new scoreboard — and it is not unanimous. PCMag handed ChatGPT its Editors' Choice award with an "Excellent" 4.0/5, while readers voted Claude their favourite at 3.5/5. Star ratings are useful, but for European users the real question is where these models actually beat each other in daily production work and how they hold up under the EU's new mandatory AI transparency rules.

September 10, 2026 Daniel Cesak

What €114.7M in EU aid for Ceuta means for AI at Europe's borders

The European Commission approved €114.7 million in emergency aid for Ceuta this week. The money buys thermal cameras, buoys and underwater drones — and the moment Europe's AI Act meets a real border.

September 9, 2026 Miriam Česáková

Open-weight AI beats GPT-4o-mini on chemistry QA – and can stay inside EU data centres

A coalition paper from Argonne, NVIDIA and PsiQuantum maps the 'convergent laboratory'. We compare its open-weight AI score with GPT-4o-mini, add EU privacy checks and price the token bill for the stack.

September 9, 2026 Daniel Cesak

From script to colour grade: 15 AI film tools and what a finished minute really costs

AI film production has gone from tech demo to documented production workflow. A new stage-by-stage guide maps 15 tools across seven parts of the film pipeline — and Colourlab Ai 4, one of the featured finishing tools, is scheduled to enter public beta on 12 September.

September 9, 2026 Daniel Cesak

Free AI tools for developers in 2026: local LLMs that keep your code private

Ollama on your own GPU, open-source coding assistants, and EU-hosted free tiers: the pragmatic privacy-first AI stack for developers in 2026 — from the workbench where we actually run it.

September 9, 2026 Daniel Cesak

Your guide to the world of artificial intelligence

Jarvis AI Magazine brings a daily overview of developments in AI – from the latest research papers to practical guides and tool reviews. We follow the development of large language models such as GPT-4o, Claude, Gemini and open-source alternatives like Llama and Mistral.

Our analyses are aimed at developers, product managers and enthusiasts who want to understand how AI is changing their field. We write clearly and without unnecessary jargon.

Don't miss any article – subscribe to our newsletter or follow the magazine regularly. Subscribe to newsletter

First, define the word before you buy the hype

Your guide to the world of artificial intelligence

Don't miss out!