AI

Claude services from Anthropic — the world's most valuable AI startup — suffered a major system-wide outage on June 23, taking down the web interface, developer API, Claude Code, and enterprise platform for over two hours. The status page reveals a troubling trend: similar incidents have occurred almost daily throughout June, and problems persist today.

Traditional automation, based on fixed rules, is hitting its limits. Get ready for the era of AI agents: they don't just follow instructions, they achieve real goals. What does this mean for Czech businesses?

Chinese lab Zhipu AI released GLM-5.2 with a full 1M-token context window and MIT license. It keeps pace with Claude Fable 5 in coding benchmarks at 25x lower cost. Elon Musk has already exchanged opinions with them.

Snyk unveils Evo Agentic Development Security (ADS) — a groundbreaking system that, for the first time, directly governs the behavior of AI agents during development. This comes as data from 10,000 environments reveals that half of developers are deploying agents directly into production, with a staggering one in twelve harboring a critical vulnerability.

Former engineers from Robinhood and WeWork have just launched Momentic — the world's first agent-based software testing platform. Here's the game-changer: AI agents don't just write tests; they fix them autonomously and monitor every single code change. Why is this crucial? Since AI programming took off, software incidents have skyrocketed by a staggering 58%.

OpenAI issues a chilling warning: Chinese entities are reportedly leveraging ChatGPT to manipulate public opinion in the US. From heated political debates to ecological controversies, AI is now fundamentally reshaping the geopolitical battlefield. But what does this mean for *your* security as a user in the EU?

NVIDIA announced the construction of 35 AI supercomputers across Europe with a combined performance of 800 exaflops. More than 3 million researchers will gain access to infrastructure for climate modeling, drug development, and quantum simulations.

Researchers have uncovered a disturbing trend: up to half of annotators on platforms like Amazon Mechanical Turk are using ChatGPT to complete assigned tasks. This creates a dangerous feedback loop, which scientists call "model collapse" — AI is training on its own outputs, and each new generation of models is worse than the last.

TikTok just unveiled its Symphony Agent and Custom Creator Networks at Cannes Lions! Dive in to see how AI is revolutionizing ad campaign creation and what it means for the European market and EU regulations.

Looking for a way to access top AI models like GPT-5.1 or Claude 4.1 without juggling multiple subscriptions? We're diving into Abacus AI ChatLLM, the all-in-one platform that promises to deliver for a fraction of the cost.