AI News

The GLM-5.2 model from China's Zhipu Z.ai represents strong competition for GPT-5.5. With a 1-million-token context window and open weights, it's changing the game for developers and businesses seeking efficiency.

French startup Mistral AI, dubbed the 'OpenAI of Europe', has posted its first job openings in Seoul. After securing over €2 billion in investments from Samsung, Nvidia, and ASML, it's expanding beyond Europe — and carrying the idea of European AI sovereignty with it.

With the rise of autonomous AI agents, the risk of credential misuse is growing. Aembit now enables businesses to securely integrate agents into the Microsoft ecosystem using temporary identities and managed access.

Forget simple chatbots. Databricks launched Genie One — agentic AI that doesn't just answer queries but actively orchestrates workflows thanks to deep understanding of enterprise context.

Artificial intelligence is facing an unexpected barrier. While models continue to grow smarter, global networks are starting to run out of high-quality data to train them. How will we solve the shortage of digital fuel?

DeepSeek raised $7.4 billion in its first funding round at a valuation exceeding $50 billion. Yet investors hold no voting rights — the capital flows into an entity controlled directly by the founder. In exchange for their billions, they got a five-year lock-up and zero influence over the company's direction.

A Chinese-British team has unveiled FinAcumen — a framework that teaches AI agents to learn from their own financial mistakes. An eight-billion-parameter model with smart memory beats specialized financial models and closes in on GPT and Claude. The code is open-source.

As users shift from simple chats to fully autonomous agents, giants like OpenAI and Anthropic face an unexpected challenge. The high computational demands of new models could lead to massive losses, just as the technology race accelerates.

Anthropic, the creator of Claude and the world's most valuable AI startup with a valuation nearing one trillion dollars, is sounding the alarm about a scenario where AI begins developing its own successors. The company is proposing international mechanisms to slow down AI development and granting governments the power to halt risky models. Shortly thereafter, the United States suspended access to Anthropic's own Fable 5 and Mythos 5 models.

Chinese startup Z.ai has unveiled the GLM-5.2 model. This 753-billion-parameter open-weight model dominates long-form coding tasks and, thanks to its innovative IndexShare architecture, delivers extreme efficiency even with massive context windows.