Anthropic’s Fable 5 & Mythos 5 Shutdown: In an unprecedented move, Anthropic was forced to pull its newly launched Fable 5 and Mythos 5 models just three days after their release following an "Is Informed" letter from the US Bureau of Industry and Security (BIS) invoking export controls. As of late June, the models remain offline.

G7 "Trusted Partners" Proposal: Prompted by the shutdown, G7 leaders discussed restricting access to cutting-edge US-developed AI models exclusively to "trusted partners," officially turning AI access into a core national security and diplomatic issue. French politicians warned against technological dependence, and Emmanuel Macron announced a coordinated Western AI cooperation platform.

.

European AI Sovereignty: OVHcloud (Europe’s largest cloud provider) announced it is developing its own frontier AI model to reduce reliance on the US and China. The company noted that training a frontier model now costs between €150M–€200M, down from previous €1B estimates.

Micron & Anthropic Infrastructure Deal: On June 22, Micron and Anthropic struck a strategic deal bundling co-design, multi-year HBM memory supply, and Series H equity to secure domestic AI compute infrastructure.

Shopify Opens Universal Commerce Protocol (UCP): Shopify made its UCP self-serve for all developers, allowing AI agents to automatically syndicate product catalogs to ChatGPT, Copilot, and Google Gemini.

Visa & OpenAI Tokenized Checkout: Visa integrated its tokenized credential infrastructure into ChatGPT and Codex, enabling AI agents to initiate secure, scoped checkouts with strict spending limits enforced before authorization.

Adyen’s Multi-Standard Adapter: Adyen launched an enterprise adapter that supports UCP, ACP, AP2, and Meta's AI checkout simultaneously, allowing merchants to avoid betting on a single agentic commerce standard.

SpaceX Acquires Cursor for $60B: SpaceX announced an all-stock acquisition of Anysphere (maker of the AI coding tool Cursor) for $60 billion (~15x revenue). Cursor, which generates ~150 million lines of enterprise code daily, will provide real-world coding data for xAI's Grok, while gaining access to SpaceX's Colossus compute cluster.

Z.ai Releases GLM-5.2: In a direct counter-narrative to the Fable 5 export controls, Z.ai (Zhipu AI) released GLM-5.2—a 753-billion-parameter open-weights model under an MIT license. Priced at roughly one-sixth the cost of GPT-5.5, it provides a powerful, ungateable alternative for enterprises.

OpenAI’s "AI Chemist" & LifeSciBench: OpenAI unveiled a "nearly autonomous AI chemist" for pharmaceutical research alongside LifeSciBench, a new evaluation framework built by 173 scientists to measure AI on real-world lab tasks rather than just trivia.

HSBC’s Massive Google Cloud Deal: HSBC signed a multi-year contract to deploy AI across 200+ business processes (fraud prevention, investment advisory), signaling a major industry shift from AI "proof of concept" experiments to company-wide ROI mandates.

Intel’s 18A-P Enters Risk Production: Driven by booming demand for AI-adjacent CPUs, Intel's next-generation 18A-P manufacturing process entered risk production, delivering 9% higher performance per watt.

AWS AgentCore Goes GA: At its Summit New York, AWS announced its AgentCore Managed Harness is now generally available and free (beyond underlying resource costs), alongside the new Continuum AI security platform.

Factory Rebrands to "Software Factories": The AI coding company Factory dropped its individual-developer framing, rebranding as a platform to orchestrate the entire software development lifecycle from spec to deployment.

Inference Costs Plummet: Z Lab and Modal published "DFlash," a speculative decoding method delivering 4.3x throughput gains on Qwen 3.5. Meanwhile, LangChain and Fireworks fine-tuned a 35B model to act as an evaluation judge that matches Claude Opus but is 100x cheaper to run.

Pew Research Study (June 17): A major new study revealed that 49% of US adults now use AI chatbots (up from 33% in 2024), and 60% read AI-generated search summaries. However, only 29% of those users actually trust the information chatbots provide, and a majority feel AI is advancing too fast.

Meta's "AI Mode" Search: Meta launched a conversational AI search engine that mines public Facebook Groups, Reels, and Marketplace listings to answer user queries directly.