June was the month AI stopped asking and started acting. It reached for your credit card, offered to scan your face, and became the subject of a tug-of-war between governments on both sides of the Atlantic. Here are the six stories from June 2026 that everyone should know about.
1. Apple Finally Gives Siri a Brain
For years, Siri has been the assistant everyone owns and no one trusts with anything important. At its annual developer conference on 8 June, Apple set out to change that, unveiling “Siri AI” — not a tweak, but a ground-up rebuild of the voice assistant that lives on more than a billion iPhones. The difference is in what the new Siri can understand. It can now see what is on your screen and draw on your personal context, so you can ask it to pull up the hotel confirmation number buried in an old email, or find the restaurant a friend recommended in a message last week. You can talk to it back and forth in a natural conversation, brainstorm ideas out loud, and even ask it for feedback on something you have written. There is a catch for us in Europe. Apple confirmed the new Siri will reach US iPhones in English later in 2026, but the rollout in the European Union and China will come later, held back by regional regulations. It is a small but telling reminder of how differently AI now arrives depending on where you live.
2. Your AI Can Now Reach for Your Wallet
Until now, an AI assistant could recommend a product but not buy it — you still had to get out your card and check out yourself. That changed on 10 June, when Visa and OpenAI announced they had plugged Visa’s payment network directly into ChatGPT, connecting it to roughly 175 million merchant locations worldwide. The practical upshot is that an AI “agent” can now find what you asked for and complete the purchase on your behalf. Crucially, you stay in charge: you set spending caps, decide which shops are allowed, and choose when the AI must check with you before paying. Behind the scenes, Visa keeps doing what it already does — running fraud checks, handling refunds, and managing chargebacks if something goes wrong. It is a genuine shift in what these tools are for. The convenience is obvious, but so is the leap of faith: handing your card to software means the guardrails you set, and how well the companies honour them, suddenly matter a great deal.
3. Banned in Three Days — Then Asking for Your ID
One of the strangest stories of the month belonged to Anthropic, the company behind the Claude assistant. On 9 June it launched its two most powerful models yet, Fable 5 and Mythos 5. Three days later, they were gone. The reason was a US government export-control directive, issued on national-security grounds, that barred all foreign nationals from accessing the models — including Anthropic’s own non-US employees. Officials had been shown a way to “jailbreak” the system to sidestep its safety limits. Unable to reliably separate users by nationality, Anthropic switched the models off worldwide to comply, while publicly disagreeing that a narrow vulnerability justified pulling a product used by millions. Running in parallel was a second, quieter development that raised eyebrows of its own. Anthropic updated its privacy policy so that it can require a small subset of flagged users to prove who they are with a government photo ID, a live selfie, and a facial-geometry scan in order to keep their account. Together, the two stories capture how entangled cutting-edge AI has become with national security — and how much personal data the price of access may soon include.
4. Europe Eases Its AI Rules — but Bans AI Deepfake Abuse
The story closest to home came on 16 June, when the European Parliament backed a deal — part of a package known as the “Digital Omnibus” — to delay the toughest parts of the EU AI Act. Obligations for high-risk AI systems that were due to bite in August 2026 have been pushed back to late 2027. The motivation was pragmatic. Europe’s landmark AI law proved complex and, in places, hard to comply with, and businesses had pushed for more time and simpler rules. For companies across the continent, including here in Luxembourg, it means more breathing room before the heaviest compliance burdens land. But this was not a simple loosening. The same package introduces a new EU-wide ban on AI systems that generate non-consensual intimate imagery and child sexual abuse material, due to take effect in December 2026. So while the red tape slows down, the protections against the most harmful uses of AI are being tightened — a balance the EU is betting will prove more workable than the original timetable.
5. America Wants One AI Rulebook, Not Fifty
As Europe eased off, the United States moved in the opposite direction — and started fighting with itself over who should be in charge. On 2 June, a presidential executive order pushed to pull AI regulation toward Washington. Two days later, lawmakers unveiled a draft bill with an unabashed name: the “Great American AI Act.” Its headline proposal is a three-year freeze that would stop individual states from writing their own frontier-AI laws, replacing a growing patchwork of state rules with a single federal approach. Supporters argue that fifty different rulebooks are unworkable for a fast-moving technology; critics counter that a freeze would strip away protections that states have moved faster than Congress to put in place. Put side by side, June’s two regulatory stories make a striking contrast. Europe is relaxing a single, unified rulebook it already has; America is arguing over whether to have one at all. For ordinary people, the outcome of both debates will quietly shape what protections they can count on and who they can hold responsible when AI causes harm.
6. The Hidden Cost of AI: Your Power Grid
Every answer an AI gives is generated somewhere — in a data centre that draws an enormous, and rapidly rising, amount of electricity. In June that usually invisible plumbing became front-page news. On 18 June, the US energy regulator FERC issued formal orders to all six of the country’s regional grid operators, telling them either to speed up the way AI data centres connect to the power grid or to justify why they cannot. It was a clear signal that the electricity system is now straining to keep up with AI’s appetite. The knock-on effects reach further than they might first appear. How quickly data centres can plug into the grid influences how fast new AI gets built — and, potentially, what households end up paying on their energy bills. It is a reminder that the AI boom is not just a software story; it runs on very physical infrastructure that everyone shares.
What to Watch in July
Three threads are worth following as the new month unfolds. The first is Anthropic’s comeback: watch for whether Fable 5 and Mythos 5 are switched back on, and note that the company’s new identity-verification policy is set to take effect on 8 July — an early test of how users react to being asked for their ID and face. The second is Europe’s next deadline. Despite June’s delay deal, the EU AI Act’s transparency rules are still slated to apply from 2 August, so expect a busy few weeks of last-minute manoeuvring over exactly what will and will not be required. And the third is AI at the checkout. With Visa and OpenAI’s agentic payments now switched on, the coming weeks will bring the first real purchases made by AI on people’s behalf — and the first real test of whether the guardrails hold up in the wild.
