Explainer tag | eYssen AI

60 Articles

Explainer ×

Where Is AI Heading? What Is Worth Preparing For

Krisztian Eyssen, Krisztian Eyssen

Where Is AI Heading? What Is Worth Preparing For

In the spring of 2025 an industry leader said that within six months AI would be writing 90% of the code. Eighteen months later the most favourable measurement, based on self-reporting, stands at **51.9%**. The direction was right, the timetable was not — and that is the rule for this whole article. There is no forecasting here: four trends, each with a measurement and a date behind it, then four claims with no evidence — among them AGI, where GPT-6 Astra substantively rearranged the argument in September.

Beginner Everyone Explainer Business, Law & Security

Sep 13, 2026

Copyright and AI: Training Data, Output, Your Rights

Krisztian Eyssen, Krisztian Eyssen

Copyright and AI: Training Data, Output, Your Rights

The first case of its kind before the Court of Justice of the European Union was referred by a **Hungarian court** — and there is still no ruling. Germany's federal supreme court heard the training-data case in September and signalled that it may send the question to Luxembourg too. What has been decided: a Munich judgment holds that copyright is infringed even when the training happened in the United States.

Intermediate Business owner Explainer Business, Law & Security

Sep 13, 2026

Deepfake Fraud: Verify Before You Pay

Krisztian Eyssen, Krisztian Eyssen

Deepfake Fraud: Verify Before You Pay

In the 25.6-million-dollar Hong Kong case it was not the video call that started the fraud but a phishing message — the deepfake was only there to remove the doubt. And according to the latest UK figures, losses to CEO fraud fell **53%** in 2025, to a record low. Prevention works. But the rule must not be tied to an amount.

Beginner Business owner Explainer Business, Law & Security

Sep 13, 2026

Prompt Injection and Data Leaks, Seen from the Business Side

Krisztian Eyssen, Krisztian Eyssen

Prompt Injection and Data Leaks, Seen from the Business Side

In Google DeepMind's own measurement, the same model showed an 18% attack success rate on the standard test and **94.6%** when the attacker knew what the defence was. There is no complete solution — the researchers say so themselves. What there is: a handful of architectural decisions a ten-person company can also make.

Intermediate Business owner Explainer Business, Law & Security

Sep 13, 2026

Shadow AI: The Tools Your Team Already Uses

Krisztian Eyssen, Krisztian Eyssen

Shadow AI: The Tools Your Team Already Uses

In a 48,340-person survey spanning 47 countries, **6%** of employees work under a policy that bans generative AI — while 41% have **no rule at all**, and 19% do not know whether one exists. The "everyone bans it, they use it anyway" story is false. The real problem is a different one, and you can map it in a single day.

Beginner Business owner Explainer Business, Law & Security

Sep 13, 2026

GDPR and AI: What Can You Feed a Chatbot?

Krisztian Eyssen, Krisztian Eyssen

GDPR and AI: What Can You Feed a Chatbot?

The European Data Protection Board has said it plainly: a model is not automatically anonymous just because it is a model. And two national regulators, the Hungarian and the Spanish, have **not a single** published decision about an AI chatbot between them — so when someone says "our regulator has ruled on this", ask them where. What does exist: seven questions to work through in order, before you type anything in.

Intermediate Business owner Explainer Business, Law & Security

Sep 13, 2026

The EU AI Act: What Applies to You Today

Krisztian Eyssen, Krisztian Eyssen

The EU AI Act: What Applies to You Today

It entered general application six weeks ago, and it was rewritten in the meantime: the high-risk rules slipped to December 2027, the transparency duties did **not**. Two new prohibitions arrive in December. And the fact most people get wrong: the market surveillance authority is not the data protection authority.

Intermediate Business owner Explainer Business, Law & Security

Sep 13, 2026

The 10 Most Common Mistakes in an AI Rollout

Krisztian Eyssen, Krisztian Eyssen

The 10 Most Common Mistakes in an AI Rollout

Every one of the ten comes with either a documented case or a survey number — and every one of them with a concrete countermove. The point is not naming the problem, it is what you do differently on Monday morning.

Beginner Business owner Explainer Business, Law & Security

Sep 13, 2026

Do Not Swap the Tool, Redesign the Workflow

Krisztian Eyssen, Krisztian Eyssen

Do Not Swap the Tool, Redesign the Workflow

Three quarters of the high-performing companies fundamentally redesigned their workflows; among the rest it was one quarter. A threefold gap — but that is self-reporting, not causal proof. The experimental measurements, on the other hand, all point the same way: the effect of AI comes through the way work is organised, not through the properties of the tool.

Intermediate Business owner Explainer Business, Law & Security

Sep 13, 2026

Measuring AI ROI – and Why Most Companies Cannot See It

Krisztian Eyssen, Krisztian Eyssen

Measuring AI ROI – and Why Most Companies Cannot See It

In the same quarter, one survey found that 95% of companies see zero return, another that three quarters of them see a positive one. Both are executives reporting on themselves. And one rigorous measurement showed that the feeling and the reality are **40 percentage points** apart. Once you know that, you already know what you have to do differently.

Intermediate Business owner Explainer Business, Law & Security

Sep 13, 2026

Fine-Tuning: When Is It Finally Worth It?

Krisztian Eyssen, Krisztian Eyssen

Fine-Tuning: When Is It Finally Worth It?

A fine-tuned model with 350 million parameters hit a **77.6%** success rate on a tool-calling task — while a general model with 175 billion parameters managed **26%** on the same task. Five hundred times fewer parameters, three times the result. That is the real promise of fine-tuning. And meanwhile one large provider is **retiring its fine-tuning platform**: no new users are being accepted, and existing ones have until January 2027.

Advanced Developer Explainer Data, RAG & Self-Hosted AI

Sep 13, 2026

Quantization: How Big Models Fit on Small Machines

Krisztian Eyssen, Krisztian Eyssen

Quantization: How Big Models Fit on Small Machines

According to a unified 2026 measurement, four-bit quantization **cuts model size by 71%** and costs on average **0.43 percentage points** of accuracy. Essentially free. Except **not for everyone**: the same four-bit quantization also produced a **59%** collapse on long input, and the degradation is **up to five times worse in languages other than English** — measured across 26 languages, Hungarian among them.

Advanced Developer Explainer Data, RAG & Self-Hosted AI

Sep 13, 2026

1
2
3
4
5

Follow Us

Archives

Tags

Audience

Everyone Office worker Marketer Creative Business owner Developer

Category

AI Basics Safety & Risk Regulation & Policy Infrastructure & Chips Business & Deals Prompting Models & Market Writing & Content Images & Design Video & Audio Marketing & SEO Coding & Development Work & Productivity Agents & Automation Data, RAG & Self-Hosted AI Business, Law & Security

Format

How-To Explainer Comparison Tool review Recipe Template Troubleshooting Case study Glossary News

Level

Beginner Intermediate Advanced

Tool

OpenRouter Perplexity HeyGen ElevenLabs Suno Obsidian MCP n8n Ollama EYAS

Others

Odoo