Anthropic and Microsoft launch general availability of Claude on NVIDIA's new GB300 Blackwell GPUs in Azure. This enables enterprises to build more powerful autonomous AI agents with improved performance and latency.

Claude Science is an integrated workbench designed specifically for scientists, unifying tools like PubMed, Jupyter, and R in a single platform. The product aims to accelerate scientific discovery by allowing researchers to work more efficiently without constantly switching between multiple tools.

Google expands Gemini's personalized AI image generation to free users in the U.S., creating content based on interests and data from connected apps. Broadens access to advanced capabilities.

Proton launches Lumo 2.0 this week with new privacy-focused capabilities. The update expands the functionality of the company's private chatbot.

Vibe-coding platform Base44 (Wix-owned) begins rolling out its own proprietary AI model. Reflects the trend of AI startups building defensibility through custom models.

OpenAI teases a Codex hardware device launching on July 15th. Could include physical shortcuts for the AI coding tool.

Crypto exchange OKX creates a marketplace where AI agents can hire and pay each other, combining payments, identity, and reputation systems. Experimental exploration of autonomous economy.

Google has added to NotebookLM the ability to generate 60-second vertical TikTok-style clips based on uploaded sources. The feature is available for Google AI Ultra and Pro subscribers, making information synthesis faster.

Flexion Robotics, founded by ex-NVIDIA engineers, showcases humanoid robot trained for useful office work. Shows advances in AI robotics for enterprise automation.

Claude Sonnet 5 brings Opus-level performance at lower prices, with significant improvements in reasoning, tool use, and coding. The model is optimized to run autonomous tasks safely, democratizing access to agentic capabilities that previously required more expensive models.

Senator Elizabeth Warren and Representative Mary Gay Scanlon propose legislation to ban sales of health and location data to brokers, including information shared with ChatGPT or Claude. Aims to protect privacy in AI.

NVIDIA presents three workflows to improve vision AI agent accuracy using synthetic data and fine-tuning in Omniverse. Particularly useful for industrial applications in factories and manufacturing.

NVIDIA has partnered with Anthropic to integrate its BioNeMo Agent Toolkit into Claude Science, providing life sciences researchers access to GPU-accelerated computing tools. This integration enables running more sophisticated scientific workflows with improved performance.

At an event for pharmaceutical executives, biotech founders, and researchers on Tuesday, Anthropic announced Claude Science, a major new product intended to support scientific research in the same way

EquiLibre Technologies, a Prague-based AI lab founded by three ex-DeepMind researchers, is now valued at more than $500 million.

Telling an LLM that 2 + 2 = 5 is enough to make it follow forbidden instructions.

Hugging Face integrates community evaluation results (Every Eval Ever) directly into model pages. Makes it easier for developers to compare model performance with verified benchmarks.

Google has updated its image generator with Nano Banana 2 Lite, optimizing it to be faster and more cost-effective. The tool is particularly useful for creators looking to generate AI-based content more efficiently.

OpenAI has introduced GeneBench-Pro, a new benchmark that evaluates AI model performance in genomics, biology, and scientific research using complex real-world datasets. This tool will be crucial for measuring progress in scientific AI applications.

New attack techniques demonstrate that AI-powered browsers can be manipulated through simple prompts that fool the model's reasoning. The finding underscores the importance of more robust guardrails in autonomous agentic systems.

Hugging Face and IBM Research have launched ScarfBench, a specialized benchmark for evaluating AI agents in Java application migration to other frameworks. This tool is relevant for measuring agent capabilities in complex enterprise tasks.

OpenAI engineers used large-scale core dump analysis to debug rare infrastructure crashes, uncovering both a hardware fault and a long-standing software bug. The case demonstrates advanced debugging techniques in large-scale AI systems.

Palantir integrates NVIDIA's open-source Nemotron models into its new AI engine for U.S. government agencies. Highlights the importance of open-source software in national security.

OpenClaw, an open-source agentic program, is finally available on mobile platforms. Its arrival on Android and iOS expands access to agentic AI tools for mobile users, democratizing their use.

Etched, an AI chip startup, has reached a $5 billion valuation with $1 billion in contracts already booked for its inference systems. This milestone reinforces competition in the AI hardware market, challenging NVIDIA's dominance.

New report shows high-intensity AI adopters saw 10.2% headcount increase, with 12% growth in entry-level roles. Counters narrative that AI eliminates junior jobs.

WIRED investigation reveals Meta contractors posed as minors to test Gemini and ChatGPT responses on high-risk topics. Raises ethical questions about competitive AI testing.

Tidal demonetizes music identified as 100% AI-generated starting July 15th but allows distribution with a label. Seeks to balance artist protection with innovation.

EquiLibre Technologies, an AI lab founded by three ex-DeepMind researchers, has reached a valuation exceeding $500 million. The startup focuses on applying AI to quantitative trading for hedge funds.

Google announces initiatives to build a nation of AI innovators in the UK focused on productivity. Part of the company's European strategy to drive AI adoption.