AI news digest — August 2, 2026
10 items, each with its source.
EU AI Act Transparency Rules Officially Take Effect
The European Union's strict transparency mandates under Article 50 are now legally enforceable. Companies must now explicitly label AI-generated content and notify users when interacting with AI systems.
reuters.comOpenAI Unveils GPT-5.5 With Advanced Reasoning Capabilities
OpenAI has officially launched GPT-5.5, featuring significant improvements in multi-step logical reasoning and reduced hallucination rates. The model is now available to ChatGPT Plus subscribers and API developers worldwide.
openai.comGoogle DeepMind Introduces AlphaFold 4 for Complex Molecules
Researchers at Google DeepMind have published a new paper detailing AlphaFold 4, which extends predictive capabilities to complex molecular interactions and dynamic protein structures. This breakthrough is expected to significantly accelerate drug discovery pipelines.
deepmind.googleBoston Dynamics Deploys Atlas Humanoids in BMW Factories
Boston Dynamics announced a major partnership to deploy its next-generation electric Atlas humanoid robots across BMW's manufacturing plants. The robots will handle complex assembly tasks alongside human workers, marking a major milestone in industrial automation.
bostondynamics.comMicrosoft Rolls Out Autonomous AI Agents in Windows 12
Microsoft's latest Windows 12 update introduces system-level autonomous AI agents capable of executing complex multi-app workflows. The feature aims to drastically improve enterprise productivity and redefine the everyday user experience.
blogs.windows.comAnthropic Secures $5 Billion to Expand Claude Ecosystem
AI startup Anthropic has closed a massive $5 billion Series E funding round led by Amazon and Google. The capital will be used to scale compute infrastructure and develop the next iteration of the Claude model family.
techcrunch.comNvidia Reveals Rubin Architecture for Next-Gen AI Workloads
Nvidia has officially unveiled its Rubin GPU architecture, promising a massive increase in training efficiency for frontier language models. The new chips are expected to begin shipping to major cloud providers in late 2026.
nvidianews.nvidia.comAI Companies Reach Landmark Settlement in Copyright Lawsuit
A coalition of major AI developers has reached a historic settlement with leading publishers regarding the use of copyrighted training data. The agreement establishes a new licensing framework and a standardized compensation model for content creators.
theverge.comMeta Open-Sources Llama 4 With 500 Billion Parameters
Meta has released the weights for Llama 4, its most powerful open-source model to date, featuring 500 billion parameters. The release includes strict safety guardrails and is available for both research and commercial use.
ai.meta.comUS Senate Approves Bipartisan AI Safety and Innovation Act
The US Senate has passed a comprehensive bill establishing federal safety standards and mandatory testing requirements for frontier AI models. The legislation now heads to the House of Representatives, bringing the US closer to a unified AI regulatory framework.
washingtonpost.comFeed, daily deep-dive and bytes — readable offline, with push alerts for the topics you follow.