Magazine

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

1093 articles

Ilustrační obrázek

Elsevier pledges €1M a year and 4,000 free AI licences for European research

Elsevier says it will put about €1 million a year into European AI research, most of it in the form of software that researchers and clinicians would otherwise have to buy. The pledge includes 4,000 free licences spread across four of the company's AI tools.

What was promised

The Dutch scientific publisher announced the commitment as part of RAISE, the European Commission's pledging initiative for AI in science.

October 15, 2026 Miriam Česáková
Ilustrační obrázek

Agility and FORT Formalise Digit 5 Safety Stack Ahead of EU Machinery Rules

Agility Robotics and FORT Robotics announced a memorandum of understanding on 1 October 2026 to build the safety stack for Digit 5, Agility's next humanoid. The agreement covers a handheld pendant, communications on the robot, and an off-robot bridge that connects Digit 5 to safety equipment already installed inside a facility. What it does not include: volumes, prices, or a deployment date. This is a reading of what the two companies actually committed to — and of the single field-hour figure that can be checked against reality.
October 15, 2026 Daniel Cesak
Ilustrační obrázek

Digit 5: Agility's humanoid drops the safety cage, lifts 22.7 kg, costs $200,000

Agility Robotics’ Digit 5 is not the first humanoid to promise a fence-free warehouse, but it is one of the first to attach hard numbers to the promise: a $200,000 machine with a 90-minute runtime, a 9-minute recharge and a claimed 91% duty cycle. For a European logistics operator, the real story is the five-year cost — roughly €370,000 per unit — and whether that survives a real picking floor.

The arithmetic behind 20+ hours a day

Agility is not promising a general-purpose domestic helper.

October 15, 2026 Daniel Cesak
Ilustrační obrázek

Nvidia Splits $2.25 Billion Between European Startups and UK AI Safety Labs

Nvidia has added $2 billion to its European spending plans and set aside $250 million of computing power for UK research into AI agent safety. One of those figures is equity money for startups. The other is five years of GPU time for the wider UK research community, with three labs named as beneficiaries of a share. Nobody has published which GPUs, or how the time gets divided.
October 15, 2026 Daniel Cesak
Ilustrační obrázek

European small firms use AI agents eight times more than in May, but the EU gap stays wide

OpenAI's economic research team reports that weekly active small-business users of its agent tools in the EU grew 8.1 times between May and October 2026. The same dataset puts EU companies at about 10% of global high-intensity AI business users, against 65% in North America.

Two numbers, one dataset, opposite moods. Agent use inside European small and medium-sized companies is climbing fast. The share of European firms that use AI heavily stays small and is not moving at the same speed.

October 15, 2026 Daniel Cesak
Ilustrační obrázek

DeepMind gives 200 European researchers its science models and $3M in compute

Google DeepMind has opened five of its scientific models to a maximum of 200 European researchers. The package covers AlphaFold 3, AlphaGenome, WeatherNext, AlphaEvolve and AlphaEarth Foundations, and comes with $3 million in Google Cloud credits. The vehicle is RAISE, an EU pilot with €107 million from Horizon Europe behind it. Per seat, the compute credit works out to about $15,000.

What was actually opened

The five models cover different fields.

October 15, 2026 Daniel Cesak
Ilustrační obrázek

TD Bank cuts a 15-hour mortgage review to under 3 minutes with agentic AI

TD Bank's in-house AI unit Layer 6 has pushed agentic AI into two more stages of the mortgage lifecycle, funding and discharge, after starting with pre-adjudication. The bank's own figures put average pre-adjudication processing time at 15 hours before the system and under 3 minutes after it. Underwriters still make the credit call — and the logging layer that makes that call auditable is the part European lenders should cost out first.

What TD automated

The system comes from TD's AI research centre, Layer 6.

October 14, 2026 Daniel Cesak
Ilustrační obrázek

Agility Robotics Heads for Nasdaq: $2.5 Billion Valuation, $1.8 Million in Sales

Agility Robotics has cleared the last big paperwork step before listing on Nasdaq. According to the joint Form S-4 registration statement filed with the US Securities and Exchange Commission by the Oregon company and Churchill Capital Corp XI, the SPAC it plans to merge with, the filing was declared effective on 13 October 2026. Shareholders vote on 2 November. If the vote passes, the combined business is set to trade as AGLT.
October 14, 2026 Daniel Cesak
Ilustrační obrázek

Mashreq runs IBM Bob across 4,000 agent requests a month in banking pilot

Mashreq, a UAE-based bank, is running IBM's agentic development platform IBM Bob in a pilot that reaches into digital banking, payments and enterprise integration work. The figures come from Zawya's report on the deployment, published on 12 October 2026, and they are pilot-phase numbers reported by the customer and the vendor rather than independently measured results: 60% to 70% less time on integration design and troubleshooting, and 4,000 AI-agent requests a month through a secure gateway. The licence price of the software is not part of that report.
October 12, 2026 Daniel Cesak
Ilustrační obrázek

Anthropic's $175 Claude Architect Exam Puts 27% of the Score on Agent Orchestration

Anthropic's Claude certification is no longer one exam. Version 1.0 of the programme, published in July 2026, splits it into four tiers priced from $99 to $175, all scored against a scaled pass mark of 720 out of 1,000 and valid for 12 months. The heaviest single domain is agentic architecture and orchestration, worth 27% of the score. That domain tests the code around the model, not the model itself.
October 12, 2026 Daniel Cesak
Ilustrační obrázek

DeepMind Wants AI Rules Sized to Each Use. Europe Already Regulates That Way

Lila Ibrahim, Google DeepMind's Chief AI Readiness Officer, says AI rules should be right-sized to each application, written into products from the start, and agreed among governments rather than left to one company. Her argument lands in Europe on ground that is already prepared. The EU AI Act sorts AI by risk, and the Commission's AI Office has held binding enforcement powers since 2 August 2026.
October 11, 2026 Miriam Česáková
Ilustrační obrázek

Meta Muse's First Month: Message Access, Rogue Sales, and No EU Launch

Meta Muse is Meta's agentic assistant. It books travel, drafts mail, moves calendar entries and can pay for things, running in the background across iOS, Android, web and WhatsApp. It went live in the US and Canada in September 2026. It is not sold in the EU. In its first month, press reports described unauthorised access to messages and Marketplace sales that owners never intended to make. Both facts matter to European readers, most of whom cannot install the app at all.

The Muse architecture

Muse runs on Meta's Muse Spark model.

October 11, 2026 Daniel Cesak
Ilustrační obrázek

How recruiters use AI on your CV in 2026, and the EU rules that help you

You send forty applications and get three replies, one of them a rejection that thanks you for your interest in a company you never applied to. It is tempting to blame a machine. The machine is part of it, but not in the place most people picture. In 2026 the AI inside a European hiring process sits in several spots at once: it parses your CV, ranks it against the job ad, schedules the interview, scores the test, and often writes the ad you answered.
October 11, 2026 Miriam Česáková
Ilustrační obrázek

USA Today Sues OpenAI for $250m: What 122 Million Training Tokens Are Worth

USA Today Co. and its affiliated publishers filed a copyright suit against OpenAI on 8 October in the Southern District of New York. The complaint covers 19 newsroom titles and asks for more than $250 million. It says ChatGPT's maker copied hundreds of thousands of articles; OpenAI says training on public web text is fair use. Both claims can sit in the same courtroom, and the interesting part is the arithmetic between them.

The filing, the parties and the demand

The case is docketed as USA Today Co., Inc. v. OpenAI Foundation, No.

October 10, 2026 Daniel Cesak
Ilustrační obrázek

ChatGPT output will carry a hidden watermark under EU labelling rules

OpenAI says it is planning an invisible text-provenance mark for ChatGPT output to support compliance with the EU's transparency rules. Article 50 does not require every AI-generated text, image, audio or video published in Europe to carry a label: duties differ between providers and deployers, and the guidance does not require watermarking for outputs shorter than 200 tokens (about 150 English words) or for code snippets.

What Article 50 asks for

Article 50 of the EU AI Act deals with transparency and labelling.

October 9, 2026 Miriam Česáková
Ilustrační obrázek

Atlassian Wants AI Agents in Jira With an ID and a Permission Scope

Atlassian used its Team '26 Europe conference in Amsterdam to introduce AMP, the Agentic Multiplayer Protocol. It gives AI agents their own identity, a scoped permission set, shared context and a governance layer inside Jira, Confluence and Loom – so an agent's work appears on the board instead of vanishing into a chat window.

That is the pitch. The more interesting parts sit underneath it: a rebuilt Model Context Protocol server, live presence indicators for agents, and a token-efficiency figure Atlassian puts at up to 25% for Jira and Confluence tasks.

October 9, 2026 Daniel Cesak
Ilustrační obrázek

The Gene and Protein Databases Behind Medical AI Get €65 Million From the EU

UniProt is where a researcher looks up what a protein does. GENCODE is where they check which stretches of the human genome are actually genes. Both are free, both are read by labs worldwide, and both have spent years living from one short grant to the next. On 6 October the European Commission said it will spend €65 million from Horizon Europe on services like these, so the work behind them can be planned further ahead.

Where the €65 million goes

The package was confirmed after an amendment to the Horizon Europe work programme.

October 8, 2026 Miriam Česáková

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

X

Don't miss out!

Subscribe for the latest news and updates.