All articles
Mistral AI reinvents itself: Vibe, physics AI, and its own data centers
At its inaugural AI NOW Summit in Paris on May 28, 2026, Mistral AI announced the rebranding of Le Chat to Vibe, an expansion into industrial AI for Airbus and BMW, and a €4 billion European data center push. Revenue target: €1 billion in 2026.
NVIDIA Isaac GR00T: open humanoid platform set for October 2026
NVIDIA and Sharpa unveiled a joint humanoid reference platform at GTC Taipei. The system integrates a Unitree H2 Plus chassis, Sharpa Wave tactile hands, and a Jetson AGX Thor T5000 module into a single research-ready environment. General availability is set for October 2026.
NVIDIA Cosmos 3 Launches as World Action Model to Replace VLA Paradigm
NVIDIA unveiled Cosmos 3 at GTC Taipei -- an open-weights physical AI model with native robotic action generation. The model displaces the VLA paradigm in favor of WAM, ranking first on four open leaderboards.
Florida Sues OpenAI and Sam Altman Over Violence Linked to ChatGPT
Florida's attorney general filed the first-ever state lawsuit against OpenAI and CEO Sam Altman on June 1. The 83-page complaint links ChatGPT to the FSU mass shooting, suicides and stalking.
Apple and Gemini on iPhone: hybrid AI over local privacy
Apple is working to bring a Gemini-powered Siri to the iPhone, but complex requests will be routed to the cloud rather than processed on device. The company signed a deal with NVIDIA for Confidential Computing — encrypted AI on Google and NVIDIA's cloud GPUs. Details are expected at WWDC 2026.
Illinois Passes First US Law Mandating Independent AI Safety Audits
Illinois has passed SB 315, mandating independent third-party AI safety audits for frontier AI companies — the first such legal requirement in the US. The House approved it 110-0 and Governor JB Pritzker announced he will sign it.
AutoTTS Cuts LLM Token Usage by 69.5% with AI-Designed Reasoning Strategies
Researchers from Meta, Google, and several universities have released AutoTTS, a framework that automatically discovers optimal test-time scaling strategies for LLMs. In tests on Qwen3 and DeepSeek models it cut token consumption by 69.5% without accuracy loss, with the discovery process costing just $39.90.
Snowflake signs $6B deal with AWS for AI CPU chips
Snowflake has signed a five-year, $6 billion contract with Amazon Web Services focused on Graviton chips for AI agent workloads. The deal nearly matches Snowflake's total AWS Marketplace revenue since 2012 and signals the maturity of the enterprise AI market.
Claude Opus 4.8: four times less likely to miss bugs, dynamic workflows added
Anthropic releases Claude Opus 4.8, succeeding Opus 4.7 with improved reliability in agentic tasks, new effort controls, and a dynamic workflows feature in Claude Code. Fast mode pricing drops threefold.
YouTube auto-labels AI videos. End of creator self-regulation
YouTube is deploying automatic AI detection and labeling in videos — independent of creator declarations. Labels move directly onto the player, and files with C2PA metadata will be permanently labeled.
Visa invests in Replit to build payment infrastructure for AI agents
Visa has announced an investment in Replit and the launch of Visa Intelligent Commerce — a program defining payment infrastructure for autonomous AI agents. Central to the initiative is the Trusted Agent Protocol (TAP), which establishes authorization rules for software-initiated transactions.
MiniMax M3: sparse attention architecture and 15.6× faster decoding
MiniMax published a technical report on its M2 series and announced M3 — a model with a new sparse attention mechanism (MSA) that decodes 15.6 times faster than M2 at one-million-token context lengths. It is the first sub-quadratic architecture the company says preserves multi-hop reasoning without compromise.