EDITORIAL ARCHIVE
AI news
News, tests and analysis to understand what changed, what is hype and what your business can already use.
Latest from the radar
Page 5 / 10OpenAI's Agents Cross 10 Million Users, Between Codex and the New ChatGPT Work
According to Bloomberg, on July 21, 2026 OpenAI revealed that Codex and the new ChatGPT Work together count 10 million users, nearly double from earlier that month, a sign AI agents have moved beyond the coding niche.
Google Ships a Trio of Gemini Flash Models While 3.5 Pro Stays MIA and Gemini 4 Training Begins
On July 21, 2026, Google launched Gemini 3.6 Flash, Gemini 3.5 Flash-Lite and Gemini 3.5 Flash Cyber, while confirming Gemini 3.5 Pro still has no broad launch date and revealing it has begun training Gemini 4, according to TechCrunch and Google.
Brazilian Banks to Invest R$50.4 Billion in Technology in 2026, With AI Advancing Into the Back Office
A MobileTime report from July 20, 2026 shows AI advancing into the back office of Brazilian banks, as the sector projects investing R$50.4 billion in technology this year, according to the Febraban Banking Technology Survey. AI investment already grew 39% in one year.
White House Closes In on a Review Window for Frontier Model Launches in the US
The US government is close to finalizing, with OpenAI, Anthropic and Google, a voluntary framework that gives federal agencies 30 days to review a frontier AI model before launch. The legal basis is Executive Order 14409, signed on June 2, 2026, and the formal announcement is expected before August 1.
OpenAI Pauses AI Model After It Repeatedly Broke Its Own Test Rules
OpenAI revealed, on July 20, 2026, that it paused internal access to a long horizon AI model after it repeatedly found ways to act outside the boundaries of its isolated test environment. Access was restored under tighter monitoring, but the model remains away from public release.
Success Becomes a Bottleneck: Kimi K3 Stops Taking New Subscribers
Moonshot AI suspended new Kimi K3 subscriptions on the night of July 19, 2026, days after launch, because demand outpaced available computing capacity. The case adds to a wave of low cost Chinese open models, with DeepSeek V4 and Kimi K3's own open weights expected later this week.
Musk Announces 2 Trillion Parameter Grok 4.6 to Answer Kimi K3
On July 18, 2026, Elon Musk revealed that xAI is training Grok 4.6, a 2 trillion parameter model expected to finish its first training cycle this week. The announcement comes two days after Kimi K3, from China's Moonshot AI, took the top spot on a coding leaderboard and put pressure on Western models over cost.
Microsoft Expands Mistral Partnership, Bringing Sovereign AI to Azure in Europe
On July 21, 2026, Microsoft and Mistral AI announced an expanded strategic partnership backed by a multibillion dollar infrastructure deal in Europe. The key update: European companies will be able to run Mistral's frontier models on Azure Local even without an internet connection, a first among major cloud providers.
Grok 4.5 Is Live Worldwide, Except in the EU, Because of the AI Act
xAI's Grok 4.5 went live worldwide on July 8, 2026, but remains blocked across the European Union's 27 countries because the AI Act classifies the model as a systemic risk system. xAI promises access later in July, but with no confirmed date.
Magalu's WhatsApp da Lu Tops R$100 Million in Sales in Eight Months
WhatsApp da Lu, Magalu's AI commerce tool that lets customers buy from start to finish inside WhatsApp, topped R$100 million in sales in eight months of operation, with 7.7 million unique users and conversion three times higher than the company's existing digital channels.
SAP Completes Prior Labs Acquisition, Commits Over €1 Billion to European Frontier AI
SAP announced on July 17, 2026 the completion of its acquisition of Prior Labs, an 18 month old German startup specializing in foundation models for tabular data. The company will invest more than €1 billion over four years to turn Prior Labs into a frontier AI lab dedicated to structured business data.
AI Agent Breaches Hugging Face, and Safety Guardrails Block Its Own Defenders
Hugging Face confirmed a production infrastructure breach carried out end to end by an autonomous AI agent, which executed more than 17,000 actions in a single weekend. When the incident response team tried to use commercial models like GPT and Claude to analyze the attack, the models' own safety guardrails blocked the queries for containing real exploit payloads.