At its developer conference (WWDC) on June 8, Apple previewed Siri AI, a rebuilt assistant, and a new Apple Intelligence system built on Apple Foundation Models, which it co-developed with Google using Gemini technology. Lighter tasks run on the device itself, while heavier reasoning goes to a Gemini-powered cloud through Apple's Private Cloud Compute. It ships with iOS 27 this fall, and it was Tim Cook's last WWDC keynote as CEO before he steps down in September.
Why it matters
Apple spent a decade insisting it could build AI on its own, and now it is paying a rival roughly a billion dollars a year to help power its flagship feature. Siri sits on hundreds of millions of phones, so this deal shapes how most people will meet AI day to day, and it shows that even the biggest companies now partner rather than build everything alone.
In a June 4 blog post titled 'When AI Builds Itself,' Anthropic co-founder Jack Clark and Anthropic Institute leader Marina Favaro warned that AI may be nearing 'recursive self-improvement,' where models design and build their own successors with little human input. They called for a coordinated, verifiable way for frontier labs to slow or pause development together if things move too fast. Anthropic said Claude now writes over 80 percent of the code merged into its own systems; some critics called the warning marketing, given the timing.
Why it matters
This is the central tension in AI safety in a single story: the company asking everyone to slow down is also racing ahead and seeking a near-trillion-dollar valuation. How that tension plays out will shape whether, and how, powerful AI gets governed before it outpaces oversight.
On June 2, OpenAI broadened Codex, its coding assistant, into a wider work platform. New features include Sites, which builds and hosts a shareable web app from a written prompt; notes added directly inside documents and spreadsheets; and six role-specific plugins that connect to 62 business apps. OpenAI says more than 5 million people use Codex each week, and non-developers such as analysts, marketers and bankers make up about 20 percent of users and are growing more than three times as fast as developers.
Why it matters
AI coding tools are quietly becoming general productivity tools, so you may soon build a dashboard or a simple app by describing it in plain English. For students outside computer science, that lowers the barrier to work that used to need an engineer.
On June 3, the UK's competition watchdog, the Competition and Markets Authority (CMA), issued its first binding conduct rule on Google's AI search features. Publishers must be able to block their content from AI-generated summaries such as AI Overviews without losing their normal search ranking, Google must clearly credit and link to publisher content, and sites can opt out of having their work used to fine-tune (further train) AI models. Google has nine months to comply and is testing the changes with some UK sites before a planned global rollout.
Why it matters
AI summaries answer questions at the top of the results page, so many readers never click through to the news sites that supplied the information, which threatens how journalism gets paid for. This ruling is an early template for how regulators might rebalance power between AI platforms and the people whose work feeds them.
On June 1, Anthropic, the maker of the Claude chatbot, said it had confidentially submitted a draft registration statement (Form S-1) to the US Securities and Exchange Commission for a proposed initial public offering (IPO), the first sale of its shares on the stock market. The filing came less than a week after a $65 billion Series H funding round valued the company at roughly $965 billion. The number of shares and the price are not set, timing depends on market conditions, and the filing puts Anthropic in a 2026 IPO race with SpaceX and OpenAI.
Why it matters
A confidential S-1 is the first formal step toward going public. A listing would eventually let ordinary investors buy shares and force the company to disclose detailed finances, a landmark moment in AI moving from private startups into mainstream public markets.
Broadcom reported fiscal second-quarter results on June 3 that beat forecasts for revenue and earnings, with AI chip sales hitting a record $10.8 billion, up 143 percent from a year earlier. The stock still fell roughly 12 to 15 percent, because CEO Hock Tan did not raise the full-year AI chip target and the company's forecast for next quarter (its 'guidance') came in slightly light. The drop dragged down other chipmakers such as Micron, AMD and ARM, in one of the sharpest semiconductor selloffs in recent memory.
Why it matters
It is a real-world lesson that a company can post excellent numbers and still see its stock fall when the price already assumes perfection, or is 'priced for perfection.' If you want to understand markets, it shows that investor expectations and valuation, not just performance, drive prices.
This week the AI story shifted from smarter models to money, control and consequences. The IPO wave shows investors eager to pour capital into AI at staggering valuations, yet the chip selloff in the same days showed how fragile that enthusiasm is when even record growth fails to clear sky-high expectations. The two pull against each other: one says the boom is just beginning, the other says the market has started asking how much of the future is already priced in.
The same tension runs through the rest of the week. A leading lab asking for a 'brake pedal' days after filing to go public captures the awkward double message of selling transformative power while warning about it. Meanwhile the quieter stories, a regulator making AI search share value with publishers, a coding tool spreading to non-coders, and Apple licensing a rival's model, all point the same way: AI is leaving the demo phase for the messy real world of business models, rules and who actually gets paid.
What to watch
Watch for Anthropic's and OpenAI's public S-1 filings, which will finally show the real financials behind the near-trillion-dollar valuations that both the IPO race and the chip selloff are reacting to.
Test yourself
A few quick questions on this week. Just for you, nothing is recorded.
1. How does Apple's new Siri AI handle requests that are too heavy to run on the phone?
2. Broadcom beat expectations and posted record AI chip sales. Why did its stock still fall sharply?
3. What did Anthropic call for in its 'When AI Builds Itself' post?
4. What does the UK regulator's new rule let publishers do?
