Anthropic limits the rollout of powerful new AI tool
Artificial intelligence giant Anthropic has built what it calls a powerful new AI model – but it is limiting access to it. On Tuesday, Anthropic said the technology will only be available to a group of 40 companies, like Amazon, Microsoft, and Apple, who will use it to locate and patch up security vulnerabilities in their software. But according to Anthropic, the model, named Mythos, broke out of its “sandbox” during testing and is capable of “circumventing” the company’s safeguards. The technology is reportedly capable of bringing down parts of the internet or even national defense systems. Anthropic said it has no plans to release the new model widely, citing concerns around cybersecurity risks and potential misuse.
NATO chief to meet with Trump administration
On a scale of awkward meetings, you’d have to give this one at least an 8 out of 10, as NATO Secretary General Mark Rutte comes to the White House just days after Donald Trump called the alliance “a paper tiger.” Trump has long questioned the value of NATO to the US and, in recent years, has successfully browbeaten other members into spending more on defense. But transatlantic tensions have soared over the past month as European allies – jilted by Trump’s decision not to read them into his Iran war plans – rebuffed Washington’s demands that they help “open” the Strait of Hormuz. Rutte, however, is regarded as something of a Trump whisperer, even if his approach to the US president (whom he once called “Daddy”) strikes many Europeans as overly sycophantic. Can Rutte’s deft touch put the world’s most formidable military alliance back on firmer footing, or are the rifts only going to deepen from here on out?

















