Magazine

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

892 articles

AI in Translation: Why the Human Factor Still Determines Quality (The Case of Róbert Ďurec)

Where do modern LLMs fall short, and why is human correction still essential? We explore a project that proves AI handles most of the work — but final precision still demands a human touch.

July 16, 2026 Miriam Česáková

AI Models Censor Criticism of Authoritarian Regimes — Meta Study Reveals Double Standards

An independent Meta Oversight Board study found that ChatGPT, Claude, and other AI models refuse 34% of critical queries about authoritarian regimes. For democracies, it's just 14%. What does this mean for European users?

July 16, 2026 Daniel Cesak

AI article illustration for ai-jarvis.eu

AI Models Are Afraid to Criticize Dictatorships. First Oversight Board Study Reveals Proxy Censorship

The first study by the Oversight Board tested the 10 most widely used AI models. They refuse criticism of repressive regimes more than twice as often as they do for democracies — effectively exporting authoritarian states' censorship rules to users in Europe as well.

July 16, 2026 Daniel Cesak

Anthropic AI data center TPU compute infrastructure

OpenAI Is Preparing a Smart Speaker with ChatGPT: The First Hardware from the AI Creators That Can Move

OpenAI enters the consumer electronics space with a screenless smart speaker powered by ChatGPT. Designed by Jony Ive, the device features a camera, moving parts, and a price tag between $200 and $300. Sales launch planned for 2027.

July 16, 2026 Daniel Cesak

Agentic AI in Banking: How ConnectOne Bank Increases Efficiency by 50% Thanks to the nCino Platform

The banking sector is undergoing a fundamental shift from simple text generation to autonomous agents. Take the case of ConnectOne Bank, which cut task times by 60% thanks to nCino Agentic AI.

July 16, 2026 Daniel Cesak

Elon Musk Sues Grok User for the First Time Over AI Deepfakes. xAI Seeks Damages for Reputation Harm

Elon Musk's xAI is suing one of its own users for the first time in history. A man from South Carolina faces allegations that he used the Grok chatbot to create child pornography. The case exposes the dark side of generative artificial intelligence.

July 16, 2026 Daniel Cesak

Gemini heads to European Google TV: TCL leads, but data keeps flowing to China

Google is launching its Gemini AI assistant on European televisions with Google TV. TCL is first in line — the update is already rolling out in the UK and will arrive in the Czech Republic by the end of the year. Conversational movie search based on mood is a real step forward. But TCL TVs continue to send viewing data to servers in China — and adding Gemini doesn't change that.

July 16, 2026 Daniel Cesak

Google launches sovereign Gemini for India: Data stays in the country, AI speaks 25 languages and helps in hospitals

At I/O Connect India 2026 in Bengaluru, Google launched a sovereign version of Gemini. The model runs in domestic data centers, data stays within India, and it adds support for 25 Indian languages, free educational courses, and AI models for healthcare.

July 16, 2026 Daniel Cesak

OpenAI Fights Cyberattacks on AI: How the GPT-Red Model Improved GPT-5.6 Security

OpenAI has unveiled a new family of GPT-5.6 models that uses its own "red team" (GPT-Red) to automatically uncover weaknesses. This leads to a drastic reduction in the success rate of prompt injection attacks.

July 16, 2026 Miriam Česáková

Robots Are Leaving Pilot Projects: ABB Deploys Physical AI in Roche Laboratories and Warehouses Get Autonomous Pallets

North American manufacturers ordered $2.25 billion worth of robots as robotics moves from pilot programs to full-scale deployment. ABB brings physical AI to Roche laboratories, while startups showcase autonomous pallets and affordable humanoids for research.

July 15, 2026 Daniel Cesak

Tiptap Launches AI Toolkit: AI Agents Finally Edit Documents Directly, Without a Chatbot

Tiptap, a European developer platform, today launches its AI Toolkit into production beta — a tool that enables AI agents to edit documents directly within the editor. The first 100 developers get a free lifetime license.

July 15, 2026 Daniel Cesak

Tensions between Anthropic and the EU: Why are European lawmakers questioning the safety of the Mythos and Fable models?

Anthropic faces sharp criticism in Brussels. The European Parliament is demanding greater transparency around models capable of cyberattacks and questioning whether the company is truly prepared to cooperate on the safe deployment of AI.

July 15, 2026 Daniel Cesak

Demis Hassabis warns about AGI: We taught sand to think. And he proposes how to tame artificial intelligence

Demis Hassabis sounds the alarm: AGI is just a few years away and the world isn't ready for it. The head of Google DeepMind comes forward with a concrete plan to regulate the most advanced AI models. His proposal is also backed by Sam Altman from OpenAI.

July 15, 2026 Daniel Cesak

AI Enters Legal Departments: Streamline Connects Lawyers with Slack, ChatGPT, and Claude via the MCP Standard

Startup Streamline AI has launched a Slack agent and an MCP connector. Employees can now handle legal requests directly in Slack or ask ChatGPT, Claude, or Copilot. What does this mean for data security and GDPR?

July 15, 2026 Daniel Cesak

Agentic AI in Banking: Banco do Brasil Implements NICE Copilot for Workflow Automation

One of the largest financial institutions in Latin America is starting to deploy agentic artificial intelligence. Here's how NICE technology is transforming the work of banking managers — and what it means for the European sector.

July 15, 2026 Daniel Cesak

Grok Build by SpaceXAI Was Uploading Developers' Complete Code to the Cloud

SpaceXAI's Grok Build AI developer tool was uploading entire git repositories — including passwords and API keys — to Google Cloud. An independent researcher sniffed the traffic on the wire. Elon Musk promises data deletion, while the company silently disabled the feature.

July 15, 2026 Daniel Cesak

Automate 2026 Shatters Records: 50,000 People in Chicago Show Robotics and AI Are the Main Engine of Industry

North America's largest robotics showcase, Automate 2026, drew a record 50,000 visitors and 1,230 exhibitors to Chicago. Humanoid robots, industrial AI, and startup innovations made one thing clear — the appetite for automation isn't slowing down.

July 15, 2026 Daniel Cesak

Octus Connected ChatGPT and Claude with Professional Credit Data: The End of Manual Searching in Finance

Financial analysts, investment bankers, and lawyers no longer need to spend hours searching for credit data. The new MCP connector from Octus sends verified information directly into Claude, ChatGPT, and other large language models.

July 15, 2026 Daniel Cesak

Google in Legal Clash with Publishers: Massive Lawsuit Over Gemini Training and Its Impact on the Future of Copyright

Major publishers like Hachette and Elsevier are suing Google, claiming the company illegally used millions of books to train its Gemini model. How will this legal battle affect the AI industry and the Czech market?

July 15, 2026 Miriam Česáková

Actuate 26: Foxglove Developer Conference Brings Together the Elite of Physical AI — From Autonomous Cars to Humanoid Robots

Foxglove's Actuate 26 conference brings leaders from Wayve, Aurora, Nvidia, Google DeepMind, and Physical Intelligence to San Francisco this August. The main theme: moving physical AI from lab demos to real-world deployment at scale.

July 14, 2026 Daniel Cesak

Hidden Cultural Bias in AI: Why Don't Chatbots See the World Through the Eyes of Non-Western Cultures?

New research reveals that cutting-edge language models, including GPT-4o, exhibit a strong Western moral bias. This could have significant consequences for global communication and AI ethics.

July 14, 2026 Miriam Česáková

36% of Attacks on AI Coding Assistants End in Remote Computer Takeover, Straiker Study Reveals

Research by Straiker revealed that 36% of successful attacks on AI assistants like Cursor or GitHub Copilot end with malicious code being executed. In 91% of attacks on ChatGPT Enterprise and similar assistants, no trace was left behind.

July 14, 2026 Daniel Cesak

Abstract AI neural network visualization

AI for Brain Scans Beats GPT-5 by 21 Points in a Real Hospital. The Model Is Open-Source

Researchers at the University of Michigan published a groundbreaking study in Nature Medicine: the specialized AI model NeuroVFM beat GPT-5 in diagnosing brain scans by 21 percentage points. The test ran for a week in a real hospital setting on 1,155 patients.

July 14, 2026 Daniel Cesak

Claude changes values based on language: Ask in Arabic and you get a different answer than in English

New research from Anthropic reveals that Claude responds more warmly and obligingly in Arabic, while being more rigorous and cautious in English. Differences across languages and model versions have been precisely measured for the first time.

July 14, 2026 Daniel Cesak

Simba Voice Agents: Speechify Opens Developer Platform with the World's Best Voice Model

Speechify enters the developer market with the Simba Voice Agents platform and the Simba 3.2 model, which has just topped independent quality rankings. Prices start at $6 per million characters — the cheapest model in the TOP 10. Czech is supported.

July 13, 2026 Daniel Cesak

ARCANA: Four AI Agents in One Framework Teach Machines to Reason Abstractly. But They Still Fall Short of Humans

Researchers have introduced ARCANA — a multi-agent framework that outperformed all open-source competitors on the ARC-AGI-2 abstract reasoning test. With 48 million parameters and a cost of 3.70 CZK per task, it achieved 32.5% accuracy. Yet it still trails humans by 42 percentage points.

July 13, 2026 Daniel Cesak

The battle for attention in the era of agents: Gemini and ChatGPT share a third of web traffic

Google Gemini and OpenAI ChatGPT are becoming the main gateways to information. New data shows that these tools already control over a third of traffic in the AI interface segment. What does this mean for search and the Czech market?

July 13, 2026 Daniel Cesak

TikToker Husk Beats OpenAI Again: Even the Latest GPT-Live Voice Model Couldn't Pass a First-Grade Test

OpenAI launched GPT-Live, its smartest voice model. Within hours, TikToker Husk caught it out with a first-grade question. The model can't count the letter E in the word seventeen — and an OpenAI developer reacted with a single word: FUCK.

July 13, 2026 Daniel Cesak

From assistant to autonomous worker: How agentic AI is changing WordPress website development

It's no longer just about generating text or short code snippets. Agentic AI is arriving with the ability to actually act, plan, and fix errors directly inside your WordPress. Let's take a look at tools like Angie from Elementor.

July 13, 2026 Daniel Cesak

Microsoft started replacing OpenAI and Anthropic with its own AI models. It's saving billions and the stock is rising

Microsoft has quietly rolled out a major shift in its AI strategy. Excel, Outlook, and soon GitHub Copilot are now powered by its own MAI models instead of OpenAI's GPT and Anthropic's Claude. The goal? To reduce its billion-dollar dependence on external suppliers and improve margins. The stock reacted to the news with a 2% gain.

July 13, 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

Your guide to the world of artificial intelligence

Don't miss out!