The week of June 8 was the moment AI stopped behaving like an industry and started behaving like infrastructure. The most powerful model on the planet was launched and then yanked offline by the US government. A rocket company minted the world’s first trillionaire on the strength of its AI ambitions. And inside one of the biggest tech firms on earth, the people building the future were in open revolt. Here are the ten stories that defined the week, counting down to the biggest.
10. Open-weight and non-US labs keep pressing
The pressure from open-weight and non-US labs never let up. China’s DeepSeek V4 Pro was touted as edging out GPT-5.5 on precision, Zhipu’s GLM 5.2 and Moonshot’s Kimi K2.7 both shipped, and Xiaomi open-sourced MiMo Code, a terminal harness it claims beats Claude Code on 200-step tasks. France’s Mistral is meanwhile rumored to be raising €3 billion at a roughly €20 billion valuation, nearly double its last round.
9. The backlash against AI data centers hardens
The physical cost of the build-out turned into real political friction. Roughly $130 billion in data-center projects have been blocked by local protests this year, Amazon owned up to using 2.5 billion gallons of water to cool its facilities last year, and grid operators keep warning the compute boom could hit a power wall by 2030. One organizer described the wins as giving communities a taste of political power.
8. A price war breaks out over cheaper AI
A price war broke out over the cost of intelligence itself. Google sharply cut its budget AI subscription tier, forcing the question of whether companies will learn to love cheaper models that do the same work for a fraction of the price. Cohere piled on by open-sourcing North Mini Code, a coding agent that runs on a single H100, even if it burns roughly three times the tokens of its rivals.
7. AI’s reliability problem keeps surfacing
AI’s trust problem kept embarrassing the companies selling it. KPMG pulled a report on AI usage after most of its citations turned out to be apparently hallucinated, and a court ruled Google legally liable for false statements its AI Overviews generate. Google’s own researchers responded by proposing “faithful uncertainty,” a way to let models admit a best guess instead of confidently inventing one.
6. OpenAI accelerates its enterprise land-grab
OpenAI spent the week buying its way deeper into the enterprise. It moved to acquire Ona to give Codex persistent cloud environments for long-running agents, scaled ChatGPT Enterprise to 100,000 BBVA employees, and reached general availability on Amazon Bedrock a month after loosening its Azure exclusivity. The expansion came as state attorneys general opened an investigation into everything from its ad policies to its handling of health data.
5. Meta’s new AI unit is in open revolt
Inside Meta, the AI strategy looked closer to mutiny. Reports described the roughly 6,500-person superintelligence group as a “soul-crushing gulag,” with staff openly rejecting a companywide AI hackathon and internal meetings turning hostile. Meta is separately moving to unwind its $2 billion Manus acquisition after Beijing ordered the deal reversed.
4. Apple bets its comeback on a Gemini-powered Siri
Apple finally made its move. At WWDC 2026 it unveiled a far more capable, personal Siri that can act across apps, backed by a new on-device foundation model and a partnership that quietly runs some models on Google’s Gemini while claiming to keep data private. The catch: Apple warned EU users may not get the features, blaming the Digital Markets Act.
3. Google out-ships everyone on raw research
Google had the busiest research week of anyone. Its DiffusionGemma open model borrows the parallel-refinement trick from image generators to write text up to four times faster, generating 256 tokens at once and self-correcting as it goes. The company also launched Gemini 3.5 Live Translate for near-instant voice-to-voice translation that preserves a speaker’s tone, plus Gemma 4, a multimodal model built to run agents locally on a laptop.
2. The AI IPO wave mints the first trillionaire
The AI money wave reshaped the markets. OpenAI confidentially filed to go public a week after Anthropic took the same step, and SpaceX’s blockbuster debut closed up 19% and delivered the world’s first trillionaire in Elon Musk, with investors now valuing the rocket maker for its AI potential. Analysts already have a name for the new cohort: it’s not FAANG anymore, it’s MANGOS.
1. Anthropic launched its best models, then the government pulled them offline
The biggest story of the week was a whiplash. On Tuesday Anthropic released Claude Fable 5 and Mythos 5, calling Fable its most powerful generally available model ever. By the weekend the company said the Commerce Department had ordered it to pull both offline worldwide on national-security grounds, after Amazon security research and reported fears of Chinese access reached the White House. Anthropic publicly disputed that a narrow jailbreak justified recalling a model used by hundreds of millions, but it complied anyway, cutting off even its own foreign employees.
Seven days that showed AI is now big enough to move stock markets, strain power grids, and get pulled offline by a government with a single order. This was the week the technology stopped being just software.