'We'll find way to save Jodrell Bank,' Burnham says Prime Minister Andy Burnham has vowed to save the UK's largest telescope at the Jodrell Bank Observatory. The Lovell telescope, which has stood in Cheshire since 1957, is the world's third-largest steerable radio telescope. But in July this year, UK Research and Innovation (UKRI) said the site would not be prioritised for further funding after its current agreement ended in March 2028. Burnham said: Every one of my generation went on a school trip to Jodrell Bank, and there is no way that we can see its demise, and it won't happen under my government. UKRI said its decision for the site was the result of a comprehensive, evidence-based prioritisation process.

Germany has said Russia was responsible for last month's drone attack at Leipzig's airport. A drone carrying an explosive device was discovered at Leipzig/Halle Airport in eastern Germany, close to Ukrainian cargo planes, on 4 August. It failed to explode as the detonator was apparently faulty, and investigators suspect a second drone collided with a cargo aircraft nearby. A third drone was reportedly found near the airport 10 days later. Interior Minister Alexander Dobrindt said the Leipzig airport attack followed the well-established pattern of Russian hybrid operations in Europe and the assumption was that Germany would be targeted again.

The vast amount of data required to model wildfires at a larger scale presents difficulties in the collection and curation of the data.

In the three years since ChatGPT's explosive debut, OpenAI's technology has upended a remarkable range of everyday activities at home, at work, and in schools. Now OpenAI is making an explicit play for scientists. In October, the firm announced that it had launched a whole new team, called OpenAI for Science, dedicated to exploring how its large language models could help scientists and tweaking its tools to support them. How does a push into science fit with OpenAI's wider mission? And what exactly is the firm hoping to achieve? I put these questions to Kevin Weil, a vice president at OpenAI who leads the new OpenAI for Science team, in an exclusive interview.

In this week's newsletter: Stranger Things' climactic showdown is the latest pop culture spectacle to feel like its been ported straight from a console. The industries' reciprocally influential relationship can be to everyone's gain Don't get Pushing Buttons delivered to your inbox? I t had begun to feel like an endurance test by the end, but nonetheless, like the sucker I am, I watched the Stranger Things finale last week. Because approximately 80% of the final season comprised twentysomething "teenagers" explaining things to each other while using random 1980s objects to illustrate convoluted plans and plot points, my expectations were not high. After an interminable hour, finally, something fun happens, as the not-kids arm themselves with machine guns and molotovs and face off against a monstrously gigantic demon-crab.

Approximate Bayesian inference for neural networks is considered a robust alternative to standard training, often providing good performance on out-of-distribution data. However, Bayesian neural networks (BNNs) with high-fidelity approximate inference via full-batch Hamiltonian Monte Carlo achieve poor generalization under covariate shift, even underperforming classical estimation. We explain this surprising result, showing how a Bayesian model average can in fact be problematic under covariate shift, particularly in cases where linear dependencies in the input features cause a lack of posterior contraction. We additionally show why the same issue does not affect many approximate inference procedures, or classical maximum a-posteriori (MAP) training. Finally, we propose novel priors that improve the robustness of BNNs to many sources of covariate shift.

About 40% of 13-to 17-year-olds in England and Wales affected by youth violence are turning to AI chatbots for mental health support. About 40% of 13-to 17-year-olds in England and Wales affected by youth violence are turning to AI chatbots for mental health support. 'I feel it's a friend': quarter of teenagers turn to AI chatbots for mental health support It was after one friend was shot and another stabbed, both fatally, that Shan asked ChatGPT for help. She had tried conventional mental health services but "chat", as she came to know her AI "friend", felt safer, less intimidating and, crucially, more available when it came to handling the trauma from the deaths of her young friends. As she started consulting the AI model, the Tottenham teenager joined about 40% of 13-to 17-year-olds in England and Wales affected by youth violence who are turning to AI chatbots for mental health support, according to research among more than 11,000 young people.

Large Language Models (LLMs) are rapidly evolving into autonomous agents capable of interacting with the external world, significantly expanding their capabilities through standardized interaction protocols. However, this paradigm revives the classic cybersecurity challenges of agency and authorization in a novel and volatile context. As decision-making shifts from deterministic code logic to probabilistic inference driven by natural language, traditional security mechanisms designed for deterministic behavior fail. It is fundamentally challenging to establish trust for unpredictable AI agents and to enforce the Principle of Least Privilege (PoLP) when instructions are ambiguous. Despite the escalating threat landscape, the academic community's understanding of this emerging domain remains fragmented, lacking a systematic framework to analyze its root causes. This paper provides a unifying formal lens for agent-interaction security. We observed that most security threats in this domain stem from a fundamental mismatch between trust evaluation and authorization policies. We introduce a novel risk analysis model centered on this trust-authorization gap. Using this model as a unifying lens, we survey and classify the implementation paths of existing, often seemingly isolated, attacks and defenses. This new framework not only unifies the field but also allows us to identify critical research gaps. Finally, we leverage our analysis to suggest a systematic research direction toward building robust, trusted agents and dynamic authorization mechanisms.

When you've finished playing the big-name video games, try,,, and some of our other favorites from 2025. It's hard to keep track of every game launch. While a handful of titles like,, or are sure to top the year's Best Of lists, many more will go unrecognized for their brilliance, fun, or sheer absurdity. The good news is we've got you covered. Whether you're stuck at home for the holidays and itching for something to play, or you just want to make sure you don't let any hidden gems slip under your radar, here are five games from this year's slate you should not miss.

AI startup Anthropic's recent announcement that it detected the world's first artificial intelligence-led hacking campaign has prompted a multitude of responses from cybersecurity experts. In a report on Friday, Anthropic said its assistant Claude Code was manipulated to carry out 80-90 percent of a "large-scale" and "highly sophisticated" cyberattack, with human intervention required "only sporadically". Anthropic, the creator of the popular Claude chatbot, said the attack aimed to infiltrate government agencies, financial institutions, tech firms and chemical manufacturing companies, though the operation was only successful in a small number of cases. The San Francisco-based company, which attributed the attack to Chinese state-sponsored hackers, did not specify how it had uncovered the operation, nor identify the "roughly" 30 entities that it said had been targeted. Roman V Yampolskiy, an AI and cybersecurity expert at the University of Louisville, said there was no doubt that AI-assisted hacking posed a serious threat, though it was difficult to verify the precise details of Anthropic's account.