AI is growing super fast! New tech, government rules, and real-world uses are popping up everywhere. Here are the most important AI news updates from this week!
Hot New Tech
OpenAI released GPT‑4.5 (based on the GPT‑4-turbo/o4 series) in February 2025, featuring faster response, better reasoning, and multimodal input (text, image, audio). It’s available in ChatGPT Plus/Pro and the OpenAI API.
GPT-5, still in development, aims to unify reasoning models with generative AI, improve causal reasoning, and learn efficiently from few examples. It is expected to launch in late 2025 or early 2026, according to Sam Altman (CEO of OpenAI).
OpenAI’s original Codex (built on GPT-3.5) powered GitHub Copilot.
Now, Codex has evolved — newer models like GPT-4 and o-series models (e.g., o1, o3, o3-pro) outperform Codex in logic and code generation.
These models now power tools like Copilot X and AI-based IDEs, making them essential for modern software engineering.
- DeepSeek & ai.com Mystery:
In early 2025, ai.com — once redirecting to ChatGPT — briefly pointed to the Chinese AI company DeepSeek.
The move caused confusion and speculation, but DeepSeek stated it was a technical misconfiguration, not a hostile domain takeover.
The redirect was reverted quickly, and no official collaboration or product launch followed.
Money & Business
- EU Supports AI with €200 Billion by 2030: The European Union is putting a lot of money (€200 billion) into AI by 2030. They want to make sure AI is used in a good way (ethical AI) and help new European AI companies grow. Europe wants to stay strong in the AI race.
- France Boosts AI with €109 Billion: France is also investing a huge amount (€109 billion) to improve its AI setup and research. Their goal is to create 50,000 new jobs in AI. They are spending a lot on data centers (places that store and process a lot of computer data) and high-performance AI (very fast and powerful AI systems).
Rules & Governments
- EU AI Act Starts: The EU AI Act (a new set of rules for AI) started its first phase on February 2, 2025. This means companies now have to give their employees mandatory AI training. Also, very strict rules for high-risk AI systems (AI that could cause serious harm) are now in place.
- AI Meeting in Paris: There was a big AI meeting in Paris with 1,500 people from 80 countries. They talked about global AI governance (how to manage AI around the world) and ethical standards (fair and good rules for AI). They all agreed on a statement about using AI responsibly.
AI Tools & How They Are Used
- Microsoft’s “AI Agents for Beginners”: Microsoft has a new, easy-to-use tool called “AI Agents for Beginners.” It helps people create their own AI assistants without needing to write any computer code. This tool uses GPT-4 and makes AI easier for everyone, even those who are not programmers.
- Azure OpenAI Service – Now Everywhere: The Azure OpenAI Service now works with many different cloud services. This means it has better security and compliance options. Businesses can now use AI models more easily and in more places.
Science & Discovery
- AI Solves Old Protein Problem: AI has solved a problem about protein folding that scientists have been trying to figure out for 18 years! It did this by using deep learning (a type of AI that learns from lots of data) and molecular dynamics (how atoms and molecules move). This is a big step for drug discovery and biotech (using living things to make products).
- More German Businesses Use AI: More and more businesses in Germany are using AI. About 20% of them are already using AI to make their processes better and improve customer service. The main challenges they face are keeping data safe and a shortage of skilled workers.
AI Safety & Good Practices
- Microsoft’s New Rules for Fair AI: Microsoft has new rules to make sure AI is fair. They focus on reducing bias (making sure AI doesn’t have unfair preferences) and fairness. They have also launched new monitoring tools to check that AI follows these good rules (ethical compliance).
- DeepSeek R1 – Better Protection Against AI Misuse: DeepSeek R1 offers stronger protection against people using AI in bad ways. It helps defend against prompt injection (tricking AI with special commands) and manipulation. It also uses federated learning (a way for AI to learn from many sources without sharing private data) to improve privacy.
What Do You Think?
Which of these new AI developments do you find most exciting? Let me know in the comments!
