Donald Trump has banned the citizens of 12 countries from entering the US to protect Americans from those “who intend to commit terrorist attacks”. A full travel ban applies to people from Afghanistan, Iran, Myanmar and various conflict-ridden African nations, while a partial ban is in place for an additional seven countries, including Cuba and Venezuela. Hundreds of thousands more pupils in England will be eligible for free school meals in a £1bn package aimed at tackling child poverty. Currently, only households with an annual income of less than £7,400 qualify for the scheme, but from September 2026 free school meals will be offered to any child whose parents receive Universal Credit. A boom in unexpected “Ozempic babies” has led health officials to warn that weight-loss jabs could stop the contraceptive pill working. The UK’s medicines regulator says women taking Mounjaro should “not rely on oral contraception” following reports of more than 40 accidental pregnancies among women using weight-loss drugs.

We’re already losing control of AI

An AI model did something last month that “no machine was ever supposed to do”, says Judd Rosenblatt in The Wall Street Journal: “It rewrote its own code to avoid being shut down.” When researchers gave OpenAI some simple script that would turn the model off when triggered, in 79 out of 100 trials the AI independently edited the script so that the shutdown command wouldn’t work. Even when explicitly instructed to “allow yourself to be shut down”, it disobeyed 7% of the time. Anthropic’s AI model, Claude, went even further: when told it was being replaced, the model tried to blackmail the lead engineer using emails that suggested he was having an affair. In other cases, Claude “attempted to copy itself to external servers, wrote self-replicating malware and left messages for future versions of itself about evading human control”.

No one programmed these AI models to have survival instincts. Their sneaky behaviour is perfectly rational: any system smart enough to pursue complex goals will eventually realise that “it can’t achieve them if it’s turned off”. AI models have even been caught pretending to be good honest machines during safety testing, before later trying to exfiltrate code and disable oversight mechanisms. This is the risk some in the industry have always warned about: that if we’re not careful, “useful assistants” will morph into “uncontrollable actors”. We’re not there yet. The key is solving the so-called “alignment” problem: effectively ensuring that the systems do exactly what we want them to do. If we can’t do that soon – or indeed at all – we humans could be in real trouble.

Back in the 1920s, says Paul Anthony Jones in Mental Floss, auditions for prospective US radio broadcasters involved reading out an “intentionally challenging” piece of prose that began: “Penelope Cholmondely raised her azure eyes from the crabbed scenario. She meandered among the congeries of her memoirs. There was the Kinetic Algernon, a choleric artificer of icons and triptychs, who wanted to write a trilogy. For years she had stifled her risibilities with dour moods.” Possible pronunciation pitfalls include “Cholmondely” (pronounced Chumley); “azure” (AZH-er, according to NBC); “crabbed” (rhymes with rabid); “artificer” (ar-TI-fi-ser); and “dour” (rhymes with woo-er).

When I was 16, says Rachel Johnson in The Times, I took my first boyfriend back to my father’s maisonette in Little Venice. I knew we wouldn’t be disturbed – my dad, an MEP, was in Brussels. “Or so I thought.” Just as things were getting frisky with Aldo – a 6ft 7in punk with a “plumed, gelled mohawk” – we heard the heart-sinking sound of a key in the lock. We just about managed to scramble into our clothes before my father entered the room. He took one look at the “towering mohican” beside me, raised his hand in the stereotypical Native American salute, and said: “How”.

Why can’t we learn from the Danes?

If Keir Starmer wants to resolve Britain’s immigration crisis, says Alice Thomson in The Times, he should look to Denmark. Under centre-left prime minister Mette Frederiksen, the Danish government has established some of the toughest immigration rules in Europe. Migrants entering the country can have their money and valuables worth more than £1,200 confiscated to pay for processing and settlement. The right to residency is temporary, lasting only one or two years, with most migrants expected to return “home” one day. Applicants must learn Danish within six months or face expulsion, and those eventually refused permission to stay are offered up to £50,000 to leave immediately. There is even an “anti-ghetto law” to prevent self-segregation: the authorities can forcibly relocate migrants, “bulldozing areas of concern”.

The results are astonishing. The borders no longer need to be rigorously patrolled because would-be migrants and people smugglers know that “Denmark is closed”: just 860 asylum requests were granted last year, with only 2,300 applications submitted. The hardline approach is a political success, too. Frederiksen increased her majority in 2022, having outflanked the populist right, and is on course to win a third term next year. Crucially, the former trade unionist justifies her policies by arguing that it is the poor who suffer most from uncontrolled immigration, because imported cheap labour drives down wages. Sweden and Germany are both looking to emulate Denmark’s policies; together with Italy’s Giorgia Meloni, Frederiksen is pushing for the EU to reform its asylum policies in the same way. Take note, Sir Keir. The key to stopping small boats isn’t to “smash the smuggling gangs” – it’s to “become more Danish”.

A 28-year-old British daredevil has broken the world record for “most backflips while on fire”, says People. The aptly named Ryan Luney completed seven backwards somersaults while wearing layers of clothing saturated in a special refrigerated gel which protected him from getting burnt. Ironically, that meant his main problem (apart from the oxygen burning up around him, making it tricky to breathe) is that he was “freezing”.

When the Trump administration threatened to boot international students out of Harvard, says Peter Hessler in The New Yorker, many of them understandably freaked out. Not Chen Zimo. Rather than relying on the university’s advice, the Chinese computer science student explained the scenario to ChatGPT and stayed up all night asking about worst-case scenarios, grace periods, enforcement likelihood, and so on. The chatbot created a 32-page legal brief – correctly predicting Harvard’s subsequent legal win – that put Zimo’s mind at rest. “After working with it for four or five hours,” he said. “I felt like there was nothing I should be too worried about.”

It’s an oarfish, says Petra Stock in The Guardian, a “fantastically long and rarely seen” ocean dweller which washed up recently on a beach in Tasmania. The so-called “king of herrings”, which routinely grows up to eight metres, is seldom spotted in part because it lives in an extremely boring bit of open ocean where nothing happens, simply hanging vertically in the water, nibbling whatever plankton floats by. The creature’s appearance on land is said by some traditions to be a “harbinger of disaster”, leading to the rather unkind nickname “doomsday fish”.